Statistics:

Definitions

Raw Data
Data collected in original form.
Frequency
The number of times a certain value or class of values occurs.
Frequency Distribution
The organization of raw data in table form with classes and frequencies.
Categorical Frequency Distribution
A frequency distribution in which the data is only nominal or ordinal.
Ungrouped Frequency Distribution
A frequency distribution of numerical data. The raw data is not grouped.
Grouped Frequency Distribution
A frequency distribution where several numbers are grouped into one class.
Class Limits
Separate one class in a grouped frequency distribution from another. The limits could
actually appear in the data and have gaps between the upper limit of one class and the
lower limit of the next.
Class Boundaries
Separate one class in a grouped frequency distribution from another. The boundaries have
one more decimal place than the raw data and therefore do not appear in the data. There
is no gap between the upper boundary of one class and the lower boundary of the next
class. The lower class boundary is found by subtracting 0.5 units from the lower class
limit and the upper class boundary is found by adding 0.5 units to the upper class limit.
Class Width
The difference between the upper and lower boundaries of any class. The class width is
also the difference between the lower limits of two consecutive classes or the upper limits
of two consecutive classes. It is not the difference between the upper and lower limits of
the same class.
Class Mark (Midpoint)
The number in the middle of the class. It is found by adding the upper and lower limits
and dividing by two. It can also be found by adding the upper and lower boundaries and
dividing by two.
Cumulative Frequency
The number of values less than the upper class boundary for the current class. This is a
running total of the frequencies.
Relative Frequency
The frequency divided by the total frequency. This gives the percent of values falling in
that class.
Cumulative Relative Frequency (Relative Cumulative Frequency)

Pharmacyst Page 1
 The running total of the relative frequencies or the cumulative frequency divided by the
total frequency. Gives the percent of the values which are less than the upper class
boundary.
Histogram
A graph which displays the data by using vertical bars of various heights to represent
frequencies. The horizontal axis can be either the class boundaries, the class marks, or the
class limits.
Frequency Polygon
A line graph. The frequency is placed along the vertical axis and the class midpoints are
placed along the horizontal axis. These points are connected with lines.
Ogive
A frequency polygon of the cumulative frequency or the relative cumulative frequency.
The vertical axis the cumulative frequency or relative cumulative frequency. The
horizontal axis is the class boundaries. The graph always starts at zero at the lowest class
boundary and will end up at the total frequency (for a cumulative frequency) or 1.00 (for
a relative cumulative frequency).
Pareto Chart
A bar graph for qualitative data with the bars arranged according to frequency.
Pie Chart
Graphical depiction of data as slices of a pie. The frequency determines the size of the
slice. The number of degrees in any slice is the relative frequency times 360 degrees.
Pictograph
A graph that uses pictures to represent data.
Stem and Leaf Plot
A data plot which uses part of the data value as the stem and the rest of the data value
(the leaf) to form groups or classes. This is very useful for sorting data quickly.

STATS: MEASURES OF CENTRAL TENDENCY

The term "Average" is vague

Average could mean one of four things. The arithmetic mean, the median, midrange, or mode.
For this reason, it is better to specify which average you're talking about.

Mean

This is what people usually intend when they say "average"

Population Mean:

Pharmacyst Page 2
Sample Mean:

Frequency Distribution:
The mean of a frequency distribution is also the weighted mean.

Median

The data must be ranked (sorted in ascending order) first. The median is the number in the
middle.
To find the depth of the median, there are several formulas that could be used, the one that we
will use is:
Depth of median = 0.5 * (n + 1)
Raw Data
The median is the number in the "depth of the median" position. If the sample size is even, the
depth of the median will be a decimal -- you need to find the midpoint between the numbers on
either side of the depth of the median.
Ungrouped Frequency Distribution
Find the cumulative frequencies for the data. The first value with a cumulative frequency greater
than depth of the median is the median. If the depth of the median is exactly 0.5 more than the
cumulative frequency of the previous class, then the median is the midpoint between the two
classes.
Grouped Frequency Distribution
This is the tough one.
Since the data is grouped, you have lost all original information. Some textbooks have you
simply take the midpoint of the class. This is an over-simplification which isn't the true value
(but much easier to do). The correct process is to interpolate.
Find out what proportion of the distance into the median class the median by dividing the sample
size by 2, subtracting the cumulative frequency of the previous class, and then dividing all that
bay the frequency of the median class.
Multiply this proportion by the class width and add it to the lower boundary of the median class.

Mode

The mode is the most frequent data value. There may be no mode if no one value appears more
than any other. There may also be two modes (bimodal), three modes (trimodal), or more than
three modes (multi-modal).

Pharmacyst Page 3
For grouped frequency distributions, the modal class is the class with the largest frequency.

Midrange

The midrange is simply the midpoint between the highest and lowest values.

Summary

The Mean is used in computing other statistics (such as the variance) and does not exist for open
ended grouped frequency distributions (1). It is often not appropriate for skewed distributions
such as salary information.
The Median is the center number and is good for skewed distributions because it is resistant to
change.
The Mode is used to describe the most typical case. The mode can be used with nominal data
whereas the others can't. The mode may or may not exist and there may be more than one value
for the mode (2).
The Midrange is not used very often. It is a very rough estimate of the average and is greatly
affected by extreme values (even more so than the mean).
Property Mean Median Mode Midrange

Always Exists No (1) Yes No (2) Yes

Uses all data values Yes No No No

Affected by extreme values Yes No No Yes

STATS: MEASURES OF VARIATION/DISPERSION

Range
The range is the simplest measure of variation to find. It is simply the highest value minus the
lowest value.
RANGE = MAXIMUM - MINIMUM
Since the range only uses the largest and smallest values, it is greatly affected by extreme values,
that is - it is not resistant to change.

Variance

Pharmacyst Page 4
"Average Deviation"
The range only involves the smallest and largest numbers, and it would be desirable to have a
statistic which involved all of the data values.
The first attempt one might make at this is something they might call the average deviation from
the mean and define it as:

The problem is that this summation is always zero. So, the average deviation will always be zero.
That is why the average deviation is never used.

Population Variance
So, to keep it from being zero, the deviation from the mean is squared and called the "squared
deviation from the mean". This "average squared deviation from the mean" is called the variance.

Unbiased Estimate of the Population Variance

One would expect the sample variance to simply be the population variance with the population
mean replaced by the sample mean. However, one of the major uses of statistics is to estimate
the corresponding parameter. This formula has the problem that the estimated value isn't the
same as the parameter. To counteract this, the sum of the squares of the deviations is divided by
one less than the sample size.

Standard Deviation

There is a problem with variances. Recall that the deviations were squared. That means that the
units were also squared. To get the units back the same as the original data values, the square
root must be taken.

The sample standard deviation is not the unbiased estimator for the population standard
deviation.

Pharmacyst Page 5
The calculator does not have a variance key on it. It does have a standard deviation key. You will
have to square the standard deviation to find the variance.

Sum of Squares (shortcuts)

The sum of the squares of the deviations from the means is given a shortcut notation and several
alternative formulas.

A little algebraic simplification returns:

What's wrong with the first formula, you ask? Consider the following example - the last row are
the totals for the columns

1. Total the data values: 23

2. Divide by the number of values to get the mean: 23/5 = 4.6
3. Subtract the mean from each value to get the numbers in the second column.
4. Square each number in the second column to get the values in the third column.
5. Total the numbers in the third column: 5.2
6. Divide this total by one less than the sample size to get the variance: 5.2 / 4 = 1.3

4 4 - 4.6 = -0.6 ( - 0.6 )^2 = 0.36

5 5 - 4.6 = 0.4 ( 0.4 ) ^2 = 0.16

3 3 - 4.6 = -1.6 ( - 1.6 )^2 = 2.56

6 6 - 4.6 = 1.4 ( 1.4 )^2 = 1.96

5 5 - 4.6 = 0.4 ( 0.4 )^2 = 0.16

23 0.00 (Always) 5.2

Not too bad, you think. But this can get pretty bad if the sample mean doesn't happen to be an
"nice" rational number. Think about having a mean of 19/7 = 2.714285714285... Those
subtractions get nasty, and when you square them, they're really bad. Another problem with the
first formula is that it requires you to know the mean ahead of time. For a calculator, this would
mean that you have to save all of the numbers that were entered. The TI-82 does this, but most
scientific calculators don't.

Pharmacyst Page 6
Now, let's consider the shortcut formula. The only things that you need to find are the sum of the
values and the sum of the values squared. There is no subtraction and no decimals or fractions
until the end. The last row contains the sums of the columns, just like before.

1. Record each number in the first column and the square of each number in the second
column.
2. Total the first column: 23
3. Total the second column: 111
4. Compute the sum of squares: 111 - 23*23/5 = 111 - 105.8 = 5.2
5. Divide the sum of squares by one less than the sample size to get the variance = 5.2 / 4 =
1.3

x x^2

4 16

5 25

3 9

6 36

5 25

23 111

Variance
Another statistical term that is related to the distribution is the variance, which is the standard
deviation squared (variance = SD² ). The SD may be either positive or negative in value because
it is calculated as a square root, which can be either positive or negative. By squaring the SD, the
problem of signs is eliminated. One common application of the variance is its use in the F-test to
compare the variance of two methods and determine whether there is a statistically significant
difference in the imprecision between the methods.
In many applications, however, the SD is often preferred because it is expressed in the same
concentration units as the data. Using the SD, it is possible to predict the range of control values
that should be observed if the method remains stable. As discussed in an earlier lesson,
laboratorians often use the SD to impose "gates" on the expected normal distribution of control
values.

Pharmacyst Page 7
Normal or Gaussian distribution

Traditionally, after
the discussion of the mean, standard deviation, degrees of freedom, and variance, the next step
was to describe the normal distribution (a frequency polygon) in terms of the standard deviation
"gates." The figure here is a representation of the frequency distribution of a large set of
laboratory values obtained by measuring a single control material. This distribution shows the
shape of a normal curve. Note that a "gate" consisting of ±1SD accounts for 68% of the
distribution or 68% of the area under the curve, ±2SD accounts for 95% and ±3SD accounts for
>99%. At ±2SD, 95% of the distribution is inside the "gates," 2.5% of the distribution is in the
lower or left tail, and the same amount (2.5%) is present in the upper tail. Some authors call this
polygon an error curve to illustrate that small errors from the mean occur more frequently than
large ones. Other authors refer to this curve as a probability distribution.

Coefficient of variation
Another way to describe the variation of a test is calculate the coefficient of variation, or CV.
The CV expresses the variation as a percentage of the mean, and is calculated as follows:
CV% = (SD/Xbar)100
In the laboratory, the CV is preferred when the SD increases in proportion to concentration. For
example, the data from a replication experiment may show an SD of 4 units at a concentration of
100 units and an SD of 8 units at a concentration of 200 units. The CVs are 4.0% at both levels
and the CV is more useful than the SD for describing method performance at concentrations in
between. However, not all tests will demonstrate imprecision that is constant in terms of CV. For
some tests, the SD may be constant over the analytical range.
The CV also provides a general "feeling" about the performance of a method. CVs of 5% or less
generally give us a feeling of good method performance, whereas CVs of 10% and higher sound
bad. However, you should look carefully at the mean value before judging a CV. At very low
concentrations, the CV may be high and at high concentrations the CV may be low. For example,
a bilirubin test with an SD of 0.1 mg/dL at a mean value of 0.5 mg/dL has a CV of 20%, whereas
an SD of 1.0 mg/dL at a concentration of 20 mg/dL corresponds to a CV of 5.0%.

Pharmacyst Page 8
Standard Error of the Mean

When we draw a random sample from a population, it is usually to infer something

about the population. Typically, from our sample we compute a statistic (such as a
sample mean) and use it to infer a population parameter (such as a population mean).
For example, let's say we run a 1 Billion hand sim and find that the initial bet
advantage (sample mean) is +1.5%. From this we usually infer that the initial bet
advantage of the entire population (population mean) of blackjack hands over all deck
subsets is 1.5%.

How accurate is the initial bet advantage (population parameter) that we find in this
way? We know that even after 1 Billion hands our measurements are approximate,
and we must give some indication of the accuracy of the measurement. This is where
the Standard Error of the Mean enters the picture.

Standard Error of the Mean:

If we were to draw all possible samples of size "n" from a given population, and for
each sample calculate the mean, and then make a frequency distribution of the sample
means, then the central limit theorem states that:

1. The mean of the distribution is the same as the mean of the population from
which the samples are drawn
2. The standard deviation of the distribution of the sample means is equal to the
standard deviation of the population divided by the square root of the sample
size "n" and is usually called the Standard Error. So,

SE = SD(of the population)/sqrt(n)

For a large sample, the distribution of the sample means is approximately a normal
distribution, even if the population from which the samples were drawn is not a
normal distribution.

Predicting the population mean from a single sample:

If the standard deviation of the means of the sample distribution, SE, is small, most of

Pharmacyst Page 9
the sample means will be near the center population mean. Thus a particular sample
mean has a good chance of being close to the population mean, and will be a good
estimator of the population mean. Conversely, a large SE means that the given sample
mean will be a poor estimator of the population mean.

However, since the frequency distribution of the sample means is normal, the chance
of a single sample mean lying within one standard deviation of the population mean is
approximately 68%. Conversely, the population mean has a 68% chance of of lying
within one standard deviation of a single randomly chosen sample mean. Thus, there
is a 68% chance that the true population mean falls within the interval:

"Sample Mean" +/- One "Standard Error"

In this way, we are able to estimate the population mean from a single sample. Not
only that, but we are able to give a range of values within which the population mean
must lie, and to also give the probability that the population mean will fall within that
interval.

There is just one difficulty: To compute the standard error SE by the central limit
theorem, we must divide the population standard deviation by the square root of the
sample size. But, the standard deviation of the population is not usually known. Thus,
it is common practice to use the sample standard deviation instead. thus the standard
error of the mean is approximated by the following formula (valid if the sample is
large):

SE = SD(of the sample)/sqrt(n)

STATS: PROBABILITY DISTRIBUTIONS

Probability Functions

A probability function is a function which assigns probabilities to the values of a random

variable.

● All the probabilities must be between 0 and 1 inclusive

● The sum of the probabilities of the outcomes must be 1.

Pharmacyst Page 10
If these two conditions aren't met, then the function isn't a probability function. There is no
requirement that the values of the random variable only be between 0 and 1, only that the
probabilities be between 0 and 1.

Probability Distributions

A listing of all the values the random variable can assume with their corresponding probabilities
make a probability distribution.
A note about random variables. A random variable does not mean that the values can be anything
(a random number). Random variables have a well defined set of outcomes and well defined
probabilities for the occurrence of each outcome. The random refers to the fact that the outcomes
happen by chance -- that is, you don't know which outcome will occur next.
Here's an example probability distribution that results from the rolling of a single fair die.
x 1 2 3 4 5 6 sum

p(x) 1/6 1/6 1/6 1/6 1/6 1/6 6/6=1

Mean, Variance, and Standard Deviation

Consider the following.

The definitions for population mean and variance used with an ungrouped frequency distribution

were:
Some of you might be confused by only dividing by N. Recall that this is the population
variance, the sample variance, which was the unbiased estimator for the population variance was
when it was divided by n-1.

Using algebra, this is equivalent to:

Recall that a probability is a long term relative frequency. So every f/N can be replaced by p(x).
This simplifies to be:
What's even better, is that the last portion of the variance is the mean squared. So, the two
formulas that we will be using are:

Here's the example we were working on earlier.

x 1 2 3 4 5 6 sum

Pharmacyst Page 11
p(x) 1/6 1/6 1/6 1/6 1/6 1/6 6/6 = 1

x p(x) 1/6 2/6 3/6 4/6 5/6 6/6 21/6 = 3.5

x^2 p(x) 1/6 4/6 9/6 16/6 25/6 36/6 91/6 = 15.1667

The mean is 7/2 or 3.5

The variance is 91/6 - (7/2)^2 = 35/12 = 2.916666...
The standard deviation is the square root of the variance = 1.7078
Do not use rounded off values in the intermediate calculations. Only round off the final answer.

Stats: Binomial Probabilities

Binomial Experiment

A binomial experiment is an experiment which satisfies these four conditions

● A fixed number of trials

● Each trial is independent of the others
● There are only two outcomes
● The probability of each outcome remains constant from trial to trial.

These can be summarized as: An experiment with a fixed number of independent trials, each of
which can only have two possible outcomes.
The fact that each trial is independent actually means that the probabilities remain constant.

Examples of binomial experiments

● Tossing a coin 20 times to see how many tails occur.

● Asking 200 people if they watch ABC news.
● Rolling a die to see if a 5 appears.

Examples which aren't binomial experiments

● Rolling a die until a 6 appears (not a fixed number of trials)

● Asking 20 people how old they are (not two outcomes)
● Drawing 5 cards from a deck for a poker hand (done without replacement, so not
independent)

Binomial Probability Function

Pharmacyst Page 12
Example:
What is the probability of rolling exactly two sixes in 6 rolls of a die?
There are five things you need to do to work a binomial story problem.

1. Define Success first. Success must be for a single trial. Success = "Rolling a 6 on a single
die"
2. Define the probability of success (p): p = 1/6
3. Find the probability of failure: q = 5/6
4. Define the number of trials: n = 6
5. Define the number of successes out of those trials: x = 2

Anytime a six appears, it is a success (denoted S) and anytime something else appears, it is a
failure (denoted F). The ways you can get exactly 2 successes in 6 trials are given below. The
probability of each is written to the right of the way it could occur. Because the trials are
independent, the probability of the event (all six dice) is the product of each probability of each
outcome (die)
1 FFFFSS 5/6 * 5/6 * 5/6 * 5/6 * 1/6 * 1/6 = (1/6)^2 * (5/6)^4
2 FFFSFS 5/6 * 5/6 * 5/6 * 1/6 * 5/6 * 1/6 = (1/6)^2 * (5/6)^4
3 FFFSSF 5/6 * 5/6 * 5/6 * 1/6 * 1/6 * 5/6 = (1/6)^2 * (5/6)^4
4 FFSFFS 5/6 * 5/6 * 1/6 * 5/6 * 5/6 * 1/6 = (1/6)^2 * (5/6)^4
5 FFSFSF 5/6 * 5/6 * 1/6 * 5/6 * 1/6 * 5/6 = (1/6)^2 * (5/6)^4
6 FFSSFF 5/6 * 5/6 * 1/6 * 1/6 * 5/6 * 5/6 = (1/6)^2 * (5/6)^4
7 FSFFFS 5/6 * 1/6 * 5/6 * 5/6 * 5/6 * 1/6 = (1/6)^2 * (5/6)^4
8 FSFFSF 5/6 * 1/6 * 5/6 * 5/6 * 1/6 * 5/6 = (1/6)^2 * (5/6)^4
9 FSFSFF 5/6 * 1/6 * 5/6 * 1/6 * 5/6 * 5/6 = (1/6)^2 * (5/6)^4
10 FSSFFF 5/6 * 1/6 * 1/6 * 5/6 * 5/6 * 5/6 = (1/6)^2 * (5/6)^4
11 SFFFFS 1/6 * 5/6 * 5/6 * 5/6 * 5/6 * 1/6 = (1/6)^2 * (5/6)^4
12 SFFFSF 1/6 * 5/6 * 5/6 * 5/6 * 1/6 * 5/6 = (1/6)^2 * (5/6)^4
13 SFFSFF 1/6 * 5/6 * 5/6 * 1/6 * 5/6 * 5/6 = (1/6)^2 * (5/6)^4
14 SFSFFF 1/6 * 5/6 * 1/6 * 5/6 * 5/6 * 5/6 = (1/6)^2 * (5/6)^4
15 SSFFFF 1/6 * 1/6 * 5/6 * 5/6 * 5/6 * 5/6 = (1/6)^2 * (5/6)^4
Notice that each of the 15 probabilities are exactly the same: (1/6)^2 * (5/6)^4.
Also, note that the 1/6 is the probability of success and you needed 2 successes. The 5/6 is the
probability of failure, and if 2 of the 6 trials were success, then 4 of the 6 must be failures. Note
that 2 is the value of x and 4 is the value of n-x.
Further note that there are fifteen ways this can occur. This is the number of ways 2 successes
can be occur in 6 trials without repetition and order not being important, or a combination of 6
things, 2 at a time.
The probability of getting exactly x success in n trials, with the probability of success on a
single trial being p is:
P(X=x) = nCx * p^x * q^(n-x)

Pharmacyst Page 13
Example:
A coin is tossed 10 times. What is the probability that exactly 6 heads will occur.

1. Success = "A head is flipped on a single coin"

2. p = 0.5
3. q = 0.5
4. n = 10
5. x=6

P(x=6) = 10C6 * 0.5^6 * 0.5^4 = 210 * 0.015625 * 0.0625 = 0.205078125

Mean, Variance, and Standard Deviation

The mean, variance, and standard deviation of a binomial distribution are extremely easy to find.

Another way to remember the variance is mu-q (since the np is mu).

Example:
Find the mean, variance, and standard deviation for the number of sixes that appear when rolling
30 dice.
Success = "a six is rolled on a single die". p = 1/6, q = 5/6.
The mean is 30 * (1/6) = 5. The variance is 30 * (1/6) * (5/6) = 25/6. The standard deviation is
the square root of the variance = 2.041241452 (approx)

Stats - Normal Distributions

Any Normal Distribution

● Bell-shaped
● Symmetric about mean
● Continuous
● Never touches the x-axis
● Total area under curve is 1.00

Pharmacyst Page 14
 ● Approximately 68% lies within 1 standard deviation of the mean, 95% within 2 standard
deviations, and 99.7% within 3 standard deviations of the mean. This is the Empirical
Rule mentioned earlier.
● Data values represented by x which has mean mu and standard deviation sigma.

● Probability Function given by

Standard Normal Distribution

Same as a normal distribution, but also ...

● Mean is zero
● Variance is one
● Standard Deviation is one
● Data values represented by z.

● Probability Function given by

Normal Probabilities

Comprehension of this table is vital to success in the course!

There is a table which must be used to look up standard normal probabilities. The z-score is
broken into two parts, the whole number and tenth are looked up along the left side and the
hundredth is looked up across the top. The value in the intersection of the row and column is the
area under the curve between zero and the z-score looked up.
Because of the symmetry of the normal distribution, look up the absolute value of any z-score.

Computing Normal Probabilities

There are several different situations that can arise when asked to find normal probabilities.
Situation Instructions

Between zero and Look up the area in the table

any number

Pharmacyst Page 15
Between two positives, or Look up both areas in the table and subtract the smaller
Between two negatives from the larger.

Between a negative and Look up both areas in the table and add them together
a positive

Less than a negative, or Look up the area in the table and subtract from 0.5000
Greater than a positive

Greater than a negative, or Look up the area in the table and add to 0.5000
Less than a positive

This can be shortened into two rules.

1. If there is only one z-score given, use 0.5000 for the second area, otherwise look up both
z-scores in the table
2. If the two numbers are the same sign, then subtract; if they are different signs, then add.
If there is only one z-score, then use the inequality to determine the second sign (< is
negative, and > is positive).

Finding z-scores from probabilities

This is more difficult, and requires you to use the table inversely. You must look up the area
between zero and the value on the inside part of the table, and then read the z-score from the
outside. Finally, decide if the z-score should be positive or negative, based on whether it was on
the left side or the right side of the mean. Remember, z-scores can be negative, but areas or
probabilities cannot be.
Situation Instructions

Area between 0 and a value Look up the area in the table

Make negative if on the left side

Area in one tail Subtract the area from 0.5000

Look up the difference in the table
Make negative if in the left tail

Area including one complete half Subtract 0.5000 from the area
(Less than a positive or greater than a Look up the difference in the table
negative) Make negative if on the left side

Pharmacyst Page 16
Within z units of the mean Divide the area by 2
Look up the quotient in the table
Use both the positive and negative z-scores

Two tails with equal area Subtract the area from 1.000
(More than z units from the mean) Divide the area by 2
Look up the quotient in the table
Use both the positive and negative z-scores

Using the table becomes proficient with practice, work lots of the normal probability problems!

STATS: CENTRAL LIMIT THEOREM

Sampling Distribution of the Sample Means

Instead of working with individual scores, statisticians often work with means. What happens is
that several samples are taken, the mean is computed for each sample, and then the means are
used as the data, rather than individual scores being used. The sample is a sampling distribution
of the sample means.
When all of the possible sample means are computed, then the following properties are true:

● The mean of the sample means will be the mean of the population
● The variance of the sample means will be the variance of the population divided by the
sample size.
● The standard deviation of the sample means (known as the standard error of the mean)
will be smaller than the population mean and will be equal to the standard deviation of
the population divided by the square root of the sample size.
● If the population has a normal distribution, then the sample means will have a normal
distribution.
● If the population is not normally distributed, but the sample size is sufficiently large, then
the sample means will have an approximately normal distribution. Some books define
sufficiently large as at least 30 and others as at least 31.

The formula for a z-score when working with the sample means is:

Finite Population Correction Factor

If the sample size is more than 5% of the population size and the sampling is done without
replacement, then a correction needs to be made to the standard error of the means.

Pharmacyst Page 17
In the following, N is the population size and n is the sample size. The adjustment is to multiply
the standard error by the square root of the quotient of the difference between the population and

sample sizes and one less than the population size.

For the most part, we will be ignoring this in class.

STATS: SAMPLE SIZE DETERMINATION

The sample size determination formulas come from the formulas for the maximum error of the
estimates. The formula is solved for n. Be sure to round the answer obtained up to the next
whole number, not off to the nearest whole number. If you round off, then you will exceed your
maximum error of the estimate in some cases. By rounding up, you will have a smaller
maximum error of the estimate than allowed, but this is better than having a larger one than
desired.

Population Mean

Here is the formula for the sample size which is obtained by solving
the maximum error of the estimate formula for the population mean
for n.

Population Proportion

Here is the formula for the sample size which is obtained by solving
the maximum error of the estimate formula for the population
proportion for n. Some texts use p hat and q hat, but since the sample
hasn't been taken, there is no value for the sample proportion. p and q
are taken from a previous study, if one is available. If there is no
previous study or estimate available, then use 0.5 for p and q, as these are the values which will
give the largest sample size, and it is better to have too large of a sample size and come under the
maximum error of the estimate than to have too small of a sample size and exceed the maximum
error of the estimate.

Parametric vs. non-parametric tests

Explanations > Social Research > Analysis > Parametric vs. non-parametric tests

Pharmacyst Page 18
There are two types of test data and consequently different types of analysis. As the table below
shows, parametric data has an underlying normal distribution which allows for more conclusions
to be drawn as the shape can be mathematically described. Anything else is non-parametric.

Parametric Non-parametric

Assumed distribution Normal Any

Assumed variance Homogeneous Any

Typical data Ratio or Interval Ordinal or Nominal

Data set relationships Independent Any

Usual central measure Mean Median

Can draw more Simplicity; Less

Benefits
conclusions affected by outliers

Tests

Choosing parametric Choosing a non-

Choosing
test parametric test

Correlation test Pearson Spearman

Independent measures, Independent-measures

Mann-Whitney test
2 groups t-test

Independent measures, One-way, independent-

Kruskal-Wallis test
>2 groups measuresANOVA

Repeated measures, 2
Matched-pair t-test
conditions Wilcoxon test

Repeated measures, >2 One-way, repeated

Friedman's test
conditions measures ANOVA

Pharmacyst Page 19
As the table shows, there are different tests for parametric and non-parametric data.

HYPOTHESIS TEST

Setting up and testing hypotheses is an essential part of statistical inference. In order to formulate
such a test, usually some theory has been put forward, either because it is believed to be true or
because it is to be used as a basis for argument, but has not been proved, for example, claiming
that a new drug is better than the current drug for treatment of the same symptoms.
In each problem considered, the question of interest is simplified into two competing claims /
hypotheses between which we have a choice; the null hypothesis, denoted H0, against the
alternative hypothesis, denoted H1. These two competing claims / hypotheses are not however
treated on an equal basis: special consideration is given to the null hypothesis.
We have two common situations:
1. The experiment has been carried out in an attempt to disprove or reject a particular
hypothesis, the null hypothesis, thus we give that one priority so it cannot be rejected
unless the evidence against it is sufficiently strong. For example,
H0: there is no difference in taste between coke and diet coke

against

H1: there is a difference.

2. If one of the two hypotheses is 'simpler' we give it priority so that a more 'complicated'
theory is not adopted unless there is sufficient evidence against the simpler one. For
example, it is 'simpler' to claim that there is no difference in flavour between coke and
diet coke than it is to say that there is a difference.

The hypotheses are often statements about population parameters like expected value and
variance; for example H0 might be that the expected value of the height of ten year old boys in
the Scottish population is not different from that of ten year old girls. A hypothesis might also be
a statement about the distributional form of a characteristic of interest, for example that the
height of ten year old boys is normally distributed within the Scottish population.
The outcome of a hypothesis test test is "Reject H0 in favour of H1" or "Do not reject H0".

NULL HYPOTHESIS

The null hypothesis, H0, represents a theory that has been put forward, either because it is
believed to be true or because it is to be used as a basis for argument, but has not been proved.
For example, in a clinical trial of a new drug, the null hypothesis might be that the new drug is
no better, on average, than the current drug. We would write

H0: there is no difference between the two drugs on average.

Pharmacyst Page 20
We give special consideration to the null hypothesis. This is due to the fact that the null
hypothesis relates to the statement being tested, whereas the alternative hypothesis relates to the
statement to be accepted if / when the null is rejected.
The final conclusion once the test has been carried out is always given in terms of the null
hypothesis. We either "Reject H0 in favour of H1" or "Do not reject H0"; we never conclude
"Reject H1", or even "Accept H1".
If we conclude "Do not reject H0", this does not necessarily mean that the null hypothesis is true,
it only suggests that there is not sufficient evidence against H0 in favour of H1. Rejecting the
null hypothesis then, suggests that the alternative hypothesis may be true.

ALTERNATIVE HYPOTHESIS

The alternative hypothesis, H1, is a statement of what a statistical hypothesis test is set up to
establish. For example, in a clinical trial of a new drug, the alternative hypothesis might be that
the new drug has a different effect, on average, compared to that of the current drug. We would
write

H1: the two drugs have different effects, on average.

The alternative hypothesis might also be that the new drug is better, on average, than the current
drug. In this case we would write

H1: the new drug is better than the current drug, on average.

The final conclusion once the test has been carried out is always given in terms of the null
hypothesis. We either "Reject H0 in favour of H1" or "Do not reject H0". We never conclude
"Reject H1", or even "Accept H1".
If we conclude "Do not reject H0", this does not necessarily mean that the null hypothesis is true,
it only suggests that there is not sufficient evidence against H0 in favour of H1. Rejecting the
null hypothesis then, suggests that the alternative hypothesis may be true.

Type I Error

In a hypothesis test, a type I error occurs when the null hypothesis is rejected when it is in fact
true; that is, H0 is wrongly rejected.
For example, in a clinical trial of a new drug, the null hypothesis might be that the new drug is
no better, on average, than the current drug; i.e.

H0: there is no difference between the two drugs on average.

A type I error would occur if we concluded that the two drugs produced different effects when in
fact there was no difference between them.

The following table gives a summary of possible results of any hypothesis test:

Pharmacyst Page 21
 Decision

Reject H0 Don't reject H0

H0 Type I Error Right decision

Truth
H1 Right decision Type II Error

A type I error is often considered to be more serious, and therefore more important to avoid, than
a type II error. The hypothesis test procedure is therefore adjusted so that there is a guaranteed
'low' probability of rejecting the null hypothesis wrongly; this probability is never 0. This
probability of a type I error can be precisely computed as

P(type I error) = significance level =

The exact probability of a type II error is generally unknown.

If we do not reject the null hypothesis, it may still be false (a type II error) as the sample may not
be big enough to identify the falseness of the null hypothesis (especially if the truth is very close
to hypothesis).
For any given set of data, type I and type II errors are inversely related; the smaller the risk of
one, the higher the risk of the other.
A type I error can also be referred to as an error of the first kind.

Type II Error

In a hypothesis test, a type II error occurs when the null hypothesis H0, is not rejected when it is
in fact false. For example, in a clinical trial of a new drug, the null hypothesis might be that the
new drug is no better, on average, than the current drug; i.e.

H0: there is no difference between the two drugs on average.

A type II error would occur if it was concluded that the two drugs produced the same effect, i.e.
there is no difference between the two drugs on average, when in fact they produced different
ones.

A type II error is frequently due to sample sizes being too small.

The probability of a type II error is generally unknown, but is symbolised by and written

P(type II error) =

A type II error can also be referred to as an error of the second kind.

Pharmacyst Page 22
Compare type I error.

Test Statistic

A test statistic is a quantity calculated from our sample of data. Its value is used to decide
whether or not the null hypothesis should be rejected in our hypothesis test.
The choice of a test statistic will depend on the assumed probability model and the hypotheses
under question.
Critical Value(s)

The critical value(s) for a hypothesis test is a threshold to which the value of the test statistic in a
sample is compared to determine whether or not the null hypothesis is rejected.
The critical value for any hypothesis test depends on the significance level at which the test is
carried out, and whether the test is one-sided or two-sided.

Critical Region

The critical region CR, or rejection region RR, is a set of values of the test statistic for which the
null hypothesis is rejected in a hypothesis test. That is, the sample space for the test statistic is
partitioned into two regions; one region (the critical region) will lead us to reject the null
hypothesis H0, the other will not. So, if the observed value of the test statistic is a member of the
critical region, we conclude "Reject H0"; if it is not a member of the critical region then we
conclude "Do not reject H0".
Significance Level

The significance level of a statistical hypothesis test is a fixed probability of wrongly rejecting
the null hypothesis H0, if it is in fact true.
It is the probability of a type I error and is set by the investigator in relation to the consequences
of such an error. That is, we want to make the significance level as small as possible in order to
protect the null hypothesis and to prevent, as far as possible, the investigator from inadvertently
making false claims.
The significance level is usually denoted by

Significance Level = P(type I error) =

Usually, the significance level is chosen to be 0.05 (or equivalently, 5%).

P-Value

Pharmacyst Page 23
The probability value (p-value) of a statistical hypothesis test is the probability of getting a value
of the test statistic as extreme as or more extreme than that observed by chance alone, if the null
hypothesis H0, is true.
It is the probability of wrongly rejecting the null hypothesis if it is in fact true.
It is equal to the significance level of the test for which we would only just reject the null
hypothesis. The p-value is compared with the actual significance level of our test and, if it is
smaller, the result is significant. That is, if the null hypothesis were to be rejected at the 5%
signficance level, this would be reported as "p < 0.05".
Small p-values suggest that the null hypothesis is unlikely to be true. The smaller it is, the more
convincing is the rejection of the null hypothesis. It indicates the strength of evidence for say,
rejecting the null hypothesis H0, rather than simply concluding "Reject H0' or "Do not reject
H0".

Power

The power of a statistical hypothesis test measures the test's ability to reject the null hypothesis
when it is actually false - that is, to make a correct decision.
In other words, the power of a hypothesis test is the probability of not committing a type II error.
It is calculated by subtracting the probability of a type II error from 1, usually expressed as:

Power = 1 - P(type II error) =

The maximum power a test can have is 1, the minimum is 0. Ideally we want a test to have high
power, close to 1.
One-sided Test

A one-sided test is a statistical hypothesis test in which the values for which we can reject the
null hypothesis, H0 are located entirely in one tail of the probability distribution.
In other words, the critical region for a one-sided test is the set of values less than the critical
value of the test, or the set of values greater than the critical value of the test.
A one-sided test is also referred to as a one-tailed test of significance.
The choice between a one-sided and a two-sided test is determined by the purpose of the
investigation or prior reasons for using a one-sided test.
Example
Suppose we wanted to test a manufacturers claim that there are, on average, 50 matches in a box.
We could set up the following hypotheses

H0: µ = 50,

against

Pharmacyst Page 24
 H1: µ < 50 or H1: µ > 50

Either of these two alternative hypotheses would lead to a one-sided test. Presumably, we would
want to test the null hypothesis against the first alternative hypothesis since it would be useful to
know if there is likely to be less than 50 matches, on average, in a box (no one would complain if
they get the correct number of matches in a box or more).

Yet another alternative hypothesis could be tested against the same null, leading this time to a
two-sided test:

H0: µ = 50,

against

H1: µ not equal to 50

Here, nothing specific can be said about the average number of matches in a box; only that, if we
could reject the null hypothesis in our test, we would know that the average number of matches
in a box is likely to be less than or greater than 50.

Two-Sided Test

A two-sided test is a statistical hypothesis test in which the values for which we can reject the
null hypothesis, H0 are located in both tails of the probability distribution.
In other words, the critical region for a two-sided test is the set of values less than a first critical
value of the test and the set of values greater than a second critical value of the test.
A two-sided test is also referred to as a two-tailed test of significance.
The choice between a one-sided test and a two-sided test is determined by the purpose of the
investigation or prior reasons for using a one-sided test.
Example
Suppose we wanted to test a manufacturers claim that there are, on average, 50 matches in a box.
We could set up the following hypotheses

H0: µ = 50,

against

H1: µ < 50 or H1: µ > 50

Either of these two alternative hypotheses would lead to a one-sided test. Presumably, we would
want to test the null hypothesis against the first alternative hypothesis since it would be useful to
know if there is likely to be less than 50 matches, on average, in a box (no one would complain if
they get the correct number of matches in a box or more).

Pharmacyst Page 25
Yet another alternative hypothesis could be tested against the same null, leading this time to a
two-sided test:

H0: µ = 50,

against

H1: µ not equal to 50

Here, nothing specific can be said about the average number of matches in a box; only that, if we
could reject the null hypothesis in our test, we would know that the average number of matches
in a box is likely to be less than or greater than 50.

One Sample t-test

A one sample t-test is a hypothesis test for answering questions about the mean where the data
are a random sample of independent observations from an underlying normal distribution
N(µ, ), where is unknown.
The null hypothesis for the one sample t-test is:

H0: µ = µ0, where µ0 is known.

That is, the sample has been drawn from a population of given mean and unknown variance
(which therefore has to be estimated from the sample).
This null hypothesis, H0 is tested against one of the following alternative hypotheses, depending
on the question posed:

H1: µ is not equal to µ

H1: µ > µ
H1: µ < µ

Two Sample t-test

A two sample t-test is a hypothesis test for answering questions about the mean where the data
are collected from two random samples of independent observations, each from an underlying
normal distribution:

When carrying out a two sample t-test, it is usual to assume that the variances for the two
populations are equal, i.e.

Pharmacyst Page 26
The null hypothesis for the two sample t-test is:
H0: µ1 = µ2

That is, the two samples have both been drawn from the same population. This null hypothesis is
tested against one of the following alternative hypotheses, depending on the question posed.
H1: µ1 is not equal to µ2
H1: µ1 > µ2
H1: µ1 < µ2

Stats: Hypothesis Testing 2:

Introduction

The first thing to do when given a claim is to write the claim mathematically (if possible), and
decide whether the given claim is the null or alternative hypothesis. If the given claim contains
equality, or a statement of no change from the given or accepted condition, then it is the null
hypothesis, otherwise, if it represents change, it is the alternative hypothesis.
The following example is not a mathematical example, but may help introduce the concept.

Example

"He's dead, Jim," said Dr. McCoy to Captain Kirk.

Mr. Spock, as the science officer, is put in charge of statistically determining the correctness of
Bones' statement and deciding the fate of the crew member (to vaporize or try to revive)
His first step is to arrive at the hypothesis to be tested.
Does the statement represent a change in previous condition?

● Yes, there is change, thus it is the alternative hypothesis, H1

● No, there is no change, therefore is the null hypothesis, H0

The correct answer is that there is change. Dead represents a change from the accepted state of
alive. The null hypothesis always represents no change. Therefore, the hypotheses are:

● H0 : Patient is alive.
● H1 : Patient is not alive (dead).

States of nature are something that you, as a statistician have no control over. Either it is, or it
isn't. This represents the true nature of things.
Possible states of nature (Based on H0)

Pharmacyst Page 27
 ● Patient is alive (H0 true - H1 false )
● Patient is dead (H0 false - H1 true)

Decisions are something that you have control over. You may make a correct decision or an
incorrect decision. It depends on the state of nature as to whether your decision is correct or in
error.
Possible decisions (Based on H0 ) / conclusions (Based on claim )

● Reject H0 / "Sufficient evidence to say patient is dead"

● Fail to Reject H0 / "Insufficient evidence to say patient is dead"

There are four possibilities that can occur based on the two possible states of nature and the two
decisions which we can make.
Statisticians will never accept the null hypothesis, we will fail to reject. In other words, we'll say
that it isn't, or that we don't have enough evidence to say that it isn't, but we'll never say that it is,
because someone else might come along with another sample which shows that it isn't and we
don't want to be wrong.

Statistically (double) speaking ...

State of Nature

Decision H0 True H0 False

Reject H0 Patient is Patient is dead,

alive,
Sufficient evidence of death
Sufficient
evidence of
death

Fail to reject H0 Patient is Patient is dead,

alive,
Insufficient evidence of death
Insufficient
evidence of
death

Which of the two errors is more serious? Type I or Type II ?

Since Type I is the more serious error (usually), that is the one we concentrate on. We usually
pick alpha to be very small (0.05, 0.01). Note: alpha is not a Type I error. Alpha is
the probability of committing a Type I error. Likewise beta is the probability of committing a
Type II error.

Pharmacyst Page 28
Conclusions
Conclusions are sentence answers which include whether there is enough evidence or not (based
on the decision), the level of significance, and whether the original claim is supported or
rejected.
Conclusions are based on the original claim, which may be the null or alternative hypotheses.
The decisions are always based on the null hypothesis
Original Claim

H0 H1
Decision "REJECT" "SUPPORT"

Reject H0 There There is sufficient evidence at the alpha

"SUFFICIENT" is sufficient evidence level of significance to support the claim
at the alpha level of that (insert original claim here)
significance
to rejectthe claim that
(insert original claim
here)

Fail to reject H0 There There is insufficient evidence at the alpha

"INSUFFICIENT" is insufficient evidence level of significance to support the claim
at the alpha level of that (insert original claim here)
significance
to rejectthe claim that
(insert original claim
here)

Stats: Type of Tests

This document will explain how to determine if the test is a left tail, right tail, or two-tail test.

Pharmacyst Page 29
The type of test is determined by
the Alternative Hypothesis ( H1 )

Left Tailed Test

H1: parameter < value

Notice the inequality points to the left

Decision Rule: Reject H0 if t.s. < c.v.

Right Tailed Test

H1: parameter > value

Notice the inequality points to the right

Decision Rule: Reject H0 if t.s. > c.v.

Two Tailed Test

H1: parameter not equal value

Another way to write not equal is < or >
Notice the inequality points to both sides

Decision Rule: Reject H0 if t.s. < c.v. (left) or t.s. > c.v. (right)

The decision rule can be summarized as follows:

Reject H0 if the test statistic falls in the critical region
(Reject H0 if the test statistic is more extreme than the critical value)

Stats: Hypothesis Testing Steps

Here are the steps to performing hypothesis testing

1. Write the original claim and identify whether it is the null hypothesis or the alternative
hypothesis.
2. Write the null and alternative hypothesis. Use the alternative hypothesis to identify
the type of test.

Pharmacyst Page 30
 3. Write down all information from the problem.
4. Find the critical value using the tables
5. Compute the test statistic
6. Make a decision to reject or fail to reject the null hypothesis. A picture showing the
critical value and test statistic may be useful.
7. Write the conclusion.

Pharmacyst Page 31
Stats: Chi-Square Distribution

The chi-square ( ) distribution is obtained from the values of

the ratio of the sample variance and population variance
multiplied by the degrees of freedom. This occurs when the
population is normally distributed with population variance
sigma^2.

Properties of the Chi-Square

● Chi-square is non-negative. Is the ratio of two non-negative values, therefore must be

non-negative itself.
● Chi-square is non-symmetric.
● There are many different chi-square distributions, one for each degree of freedom.
● The degrees of freedom when working with a single population variance is n-1.

Chi-Square Probabilities

Since the chi-square distribution isn't symmetric, the method for looking up left-tail values is
different from the method for looking up right tail values.

● Area to the right - just use the area given.

● Area to the left - the table requires the area to the right, so subtract the given area from
one and look this area up in the table.
● Area in both tails - divide the area by two. Look up this area for the right critical value
and one minus this area for the left critical value.

DF which aren't in the table

When the degrees of freedom aren't listed in the table, there are a couple of choices that you
have.

● You can interpolate. This is probably the more accurate way. Interpolation involves
estimating the critical value by figuring how far the given degrees of freedom are
between the two df in the table and going that far between the critical values in the table.
Most people born in the 70's didn't have to learn interpolation in high school because they
had calculators which would do logarithms (we had to use tables in the "good old" days).
● You can go with the critical value which is less likely to cause you to reject in error (type
I error). For a right tail test, this is the critical value further to the right (larger). For a left
tail test, it is the value further to the left (smaller). For a two-tail test, it's the value further
to the left and the value further to the right. Note, it is not the column with the degrees of
freedom further to the right, it's the critical value which is further to the right.

Pharmacyst Page 32
Stats: Goodness-of-fit Test

The idea behind the chi-square goodness-of-fit test is to see if the sample comes from the
population with the claimed distribution. Another way of looking at that is to ask if the frequency
distribution fits a specific pattern.
Two values are involved, an observed value, which is the frequency of a category from a sample,
and the expected frequency, which is calculated based upon the claimed distribution. The
derivation of the formula is very similar to that of the variance which was done earlier (chapter 2
or 3).
The idea is that if the observed frequency is really close to the claimed (expected) frequency,
then the square of the deviations will be small. The square of the deviation is divided by the
expected frequency to weight frequencies. A difference of 10 may be very significant if 12 was
the expected frequency, but a difference of 10 isn't very significant at all if the expected
frequency was 1200.
If the sum of these weighted squared deviations is small, the observed frequencies are close to
the expected frequencies and there would be no reason to reject the claim that it came from that
distribution. Only when the sum is large is the a reason to question the distribution. Therefore,
the chi-square goodness-of-fit test is always a right tail
test.
The test statistic has a chi-square distribution when the
following assumptions are met

● The data are obtained from a random sample

● The expected frequency of each category must be at least 5. This goes back to the
requirement that the data be normally distributed. You're simulating a multinomial
experiment (using a discrete distribution) with the goodness-of-fit test (and a continuous
distribution), and if each expected frequency is at least five then you can use the normal
distribution to approximate (much like the binomial). If the expected

The following are properties of the goodness-of-fit test

● The data are the observed frequencies. This means that there is only one data value for
each category. Therefore, ...
● The degrees of freedom is one less than the number of categories, not one less than the
sample size.
● It is always a right tail test.
● It has a chi-square distribution.
● The value of the test statistic doesn't change if the order of the categories is switched.

● The test statistic is

Interpreting the Claim

Pharmacyst Page 33
There are four ways you might be given a claim.

1. The values occur with equal frequency. Other words for this are "uniform", "no
preference", or "no difference". To find the expected frequencies, total the observed
frequencies and divide by the number of categories. This quotient is the expected
frequency for each category.
2. Specific proportions or probabilities are given. To find the expected frequencies, multiply
the total of the observed frequencies by the probability for each category.
3. The expected frequencies are given to you. In this case, you don't have to do anything.
4. A specific distribution is claimed. For example, "The data is normally distributed". To
work a problem like this, you need to group the data and find the frequency for each
class. Then, find the probability of being within that class by converting the scores to z-
scores and looking up the probabilities. Finally, multiply the probabilities by the total
observed frequency. (It's not really as bad as it sounds).

Stats: F-Test

Definitions
F-distribution

The ratio of two independent chi-square variables divided by their respective degrees of
freedom. If the population variances are equal, this simplifies to be the ratio of the sample
variances.

Analysis of Variance (ANOVA)

A technique used to test a hypothesis concerning the means of three or mor populations.

One-Way Analysis of Variance

Analysis of Variance when there is only one independent variable. The null hypothesis
will be that all population means are equal, the alternative hypothesis is that at least one
mean is different.

Between Group Variation

The variation due to the interaction between the samples, denoted SS(B) for Sum of
Squares Between groups. If the sample means are close to each other (and therefore the
Grand Mean) this will be small. There are k samples involved with one data value for
each sample (the sample mean), so there are k-1 degrees of freedom.

Between Group Variance

Pharmacyst Page 34
 The variance due to the interaction between the samples, denoted MS(B) for Mean
Square Between groups. This is the between group variation divided by its degrees of
freedom.

Within Group Variation

The variation due to differences within individual samples, denoted SS(W) for Sum of
Squares Within groups. Each sample is considered independently, no interaction between
samples is involved. The degrees of freedom is equal to the sum of the individual degrees
of freedom for each sample. Since each sample has degrees of freedom equal to one less
than their sample sizes, and there are k samples, the total degrees of freedom is k less
than the total sample size: df = N - k.

Within Group Variance

The variance due to the differences within individual samples, denoted MS(W) for Mean
Square Within groups. This is the within group variation divided by its degrees of
freedom.

Scheffe' Test

A test used to find where the differences between means lie when the Analysis of
Variance indicates the means are not all equal. The Scheffe' test is generally used when
the sample sizes are different.

Tukey Test

A test used to find where the differences between the means lie when the Analysis of
Variance indicates the means are not all equal. The Tukey test is generally used when the
sample sizes are all the same.

Two-Way Analysis of Variance

An extension to the one-way analysis of variance. There are two independent variables.
There are three sets of hypothesis with the two-way ANOVA. The first null hypothesis is
that there is no interaction between the two factors. The second null hypothesis is that the
population means of the first factor are equal. The third null hypothesis is that the
population means of the second factor are equal.

Factors

The two independent variables in a two-way ANOVA.

Treatment Groups

Pharmacyst Page 35
 Groups formed by making all possible combinations of the two factors. For example, if
the first factor has 3 levels and the second factor has 2 levels, then there will be 3x2=6
different treatment groups.

Interaction Effect

The effect one factor has on the other factor

Main Effect

The effects of the independent variables.

Stats: F-Test

The F-distribution is formed by the ratio of two independent chi-square

variables divided by their respective degrees of freedom.
Since F is formed by chi-square, many of the chi-square properties carry
over to the F distribution.

● The F-values are all non-negative

● The distribution is non-symmetric
● The mean is approximately 1
● There are two independent degrees of freedom, one for the numerator, and one for the
denominator.
● There are many different F distributions, one for each pair of degrees of freedom.

F-Test

The F-test is designed to test if two population variances are equal. It does this by comparing the
ratio of two variances. So, if the variances are equal, the ratio of the variances will be 1.

All hypothesis testing is done under the assumption the null hypothesis is true

If the null hypothesis is true, then the F test-statistic given above can be simplified
(dramatically). This ratio of sample variances will be test statistic used. If the null
hypothesis is false, then we will reject the null hypothesis that the ratio was equal to
1 and our assumption that they were equal.

Pharmacyst Page 36
There are several different F-tables. Each one has a different level of significance. So, find the
correct level of significance first, and then look up the numerator degrees of freedom and the
denominator degrees of freedom to find the critical value.
You will notice that all of the tables only give level of significance for right tail tests. Because
the F distribution is not symmetric, and there are no negative values, you may not simply take
the opposite of the right critical value to find the left critical value. The way to find a left critical
value is to reverse the degrees of freedom, look up the right critical value, and then take the
reciprocal of this value. For example, the critical value with 0.05 on the left with 12 numerator
and 15 denominator degrees of freedom is found of taking the reciprocal of the critical value
with 0.05 on the right with 15 numerator and 12 denominator degrees of freedom.

Avoiding Left Critical Values

Since the left critical values are a pain to calculate, they are often avoided altogether. This is the
procedure followed in the textbook. You can force the F test into a right tail test by placing the
sample with the large variance in the numerator and the smaller variance in the denominator. It
does not matter which sample has the larger sample size, only which sample has the larger
variance.
The numerator degrees of freedom will be the degrees of freedom for whichever sample has the
larger variance (since it is in the numerator) and the denominator degrees of freedom will be the
degrees of freedom for whichever sample has the smaller variance (since it is in the
denominator).
If a two-tail test is being conducted, you still have to divide alpha by 2, but you only look up and
compare the right critical value.
Assumptions / Notes

● The larger variance should always be placed in the numerator

● The test statistic is F = s1^2 / s2^2 where s1^2 > s2^2
● Divide alpha by 2 for a two tail test and then find the right critical value
● If standard deviations are given instead of variances, they must be squared
● When the degrees of freedom aren't given in the table, go with the value with the larger
critical value (this happens to be the smaller degrees of freedom). This is so that you are
less likely to reject in error (type I error)
● The populations from which the samples were obtained must be normal.
● The samples must be independent

Stats: One-Way ANOVA

Pharmacyst Page 37
A One-Way Analysis of Variance is a way to test the equality of three or more means at one time
by using variances.

Assumptions

● The populations from which the samples were obtained must be normally or
approximately normally distributed.
● The samples must be independent.
● The variances of the populations must be equal.

Hypotheses
The null hypothesis will be that all population means are equal, the alternative hypothesis is that
at least one mean is different.
In the following, lower case letters apply to the individual samples and capital letters apply to the
entire set collectively. That is, n is one of many sample sizes, but N is the total sample size.

Grand Mean
The grand mean of a set of samples is the total of all the data values divided by
the total sample size. This requires that you have all of the sample data available
to you, which is usually the case, but not always. It turns out that all that is
necessary to find perform a one-way analysis of variance are the number of samples, the sample
means, the sample variances, and the sample sizes.
Another way to find the grand mean is to find the weighted average of the
sample means. The weight applied is the sample size.

Total Variation
The total variation (not variance) is comprised the sum of the squares of the differences of each
mean with the grand mean.
There is the between group variation and the within group variation. The whole idea behind the
analysis of variance is to compare the ratio of between group variance to within group variance.
If the variance caused by the interaction between the samples is much larger when compared to
the variance that appears within each group, then it is because the means aren't the same.

Between Group Variation

The variation due to the interaction between the samples is
denoted SS(B) for Sum of Squares Between groups. If the
sample means are close to each other (and therefore the Grand Mean) this will be small. There
are k samples involved with one data value for each sample (the sample mean), so there are k-1
degrees of freedom.

Pharmacyst Page 38
The variance due to the interaction between the samples is denoted MS(B) for Mean Square
Between groups. This is the between group variation divided by its degrees of freedom. It is also

denoted by .

Within Group Variation

The variation due to differences within individual samples, denoted SS(W) for Sum of Squares
Within groups. Each sample is considered independently, no interaction between samples is
involved. The degrees of freedom is equal to the sum of the individual degrees of freedom for
each sample. Since each sample has degrees of freedom equal to one less than their sample sizes,
and there are k samples, the total degrees of freedom is k less than the total sample size: df = N -
k.
The variance due to the differences within individual samples is denoted MS(W) for Mean
Square Within groups. This is the within group variation divided by its degrees of freedom. It is

also denoted by . It is the weighted average of the variances (weighted with the degrees of
freedom).

F test statistic
Recall that a F variable is the ratio of two independent chi-square variables divided
by their respective degrees of freedom. Also recall that the F test statistic is the
ratio of two sample variances, well, it turns out that's exactly what we have here.
The F test statistic is found by dividing the between group variance by the within
group variance. The degrees of freedom for the numerator are the degrees of freedom for the
between group (k-1) and the degrees of freedom for the denominator are the degrees of freedom
for the within group (N-k).

Summary Table
All of this sounds like a lot to remember, and it is. However, there is a table which makes things
really nice.
SS df MS F

Between SS(B) k-1 SS(B) MS(B)

----------- --------------
k-1 MS(W)

Within SS(W) N-k SS(W) .

-----------

Pharmacyst Page 39
 N-k

Total SS(W) + SS(B) N-1 . .

Notice that each Mean Square is just the Sum of Squares divided by its degrees of freedom, and
the F value is the ratio of the mean squares. Do not put the largest variance in the numerator,
always divide the between variance by the within variance. If the between variance is smaller
than the within variance, then the means are really close to each other and you will fail to reject
the claim that they are all equal. The degrees of freedom of the F-test are in the same order they
appear in the table (nifty, eh?).

Decision Rule
The decision will be to reject the null hypothesis if the test statistic from the table is greater than
the F critical value with k-1 numerator and N-k denominator degrees of freedom.
If the decision is to reject the null, then at least one of the means is different. However, the
ANOVA does not tell you where the difference lies. For this, you need another test, either
the Scheffe' or Tukey test.

TI-82
Ok, now for the really good news. There's a program called ANOVA for the TI-82 calculator
which will do all of the calculations and give you the values that go into the table for you. You
must have the sample means, sample variances, and sample sizes to use the program. If you have
the sum of squares, then it is much easier to finish the table by hand (this is what we'll do with
the two-way analysis of variance)

Stats: Two-Way ANOVA

The two-way analysis of variance is an extension to the one-way analysis of variance. There are
two independent variables (hence the name two-way).

Assumptions

● The populations from which the samples were obtained must be normally or
approximately normally distributed.
● The samples must be independent.
● The variances of the populations must be equal.
● The groups must have the same sample size.

Hypotheses
There are three sets of hypothesis with the two-way ANOVA.

Pharmacyst Page 40
The null hypotheses for each of the sets are given below.

1. The population means of the first factor are equal. This is like the one-way ANOVA for
the row factor.
2. The population means of the second factor are equal. This is like the one-way ANOVA
for the column factor.
3. There is no interaction between the two factors. This is similar to performing a test for
independence with contingency tables.

Factors
The two independent variables in a two-way ANOVA are called factors. The idea is that there
are two variables, factors, which affect the dependent variable. Each factor will have two or more
levels within it, and the degrees of freedom for each factor is one less than the number of levels.

Treatment Groups
Treatement Groups are formed by making all possible combinations of the two factors. For
example, if the first factor has 3 levels and the second factor has 2 levels, then there will be
3x2=6 different treatment groups.
As an example, let's assume we're planting corn. The type of seed and type of fertilizer are the
two factors we're considering in this example. This example has 15 treatment groups. There are
3-1=2 degrees of freedom for the type of seed, and 5-1=4 degrees of freedom for the type of
fertilizer. There are 2*4 = 8 degrees of freedom for the interaction between the type of seed and
type of fertilizer.
The data that actually appears in the table are samples. In this case, 2 samples from each
treatment group were taken.
Fert I Fert II Fert III Fert IV Fert V

Seed A-402 106, 110 95, 100 94, 107 103, 104 100, 102

Seed B-894 110, 112 98, 99 100, 101 108, 112 105, 107

Seed C-952 94, 97 86, 87 98, 99 99, 101 94, 98

Main Effect
The main effect involves the independent variables one at a time. The interaction is ignored for
this part. Just the rows or just the columns are used, not mixed. This is the part which is similar
to the one-way analysis of variance. Each of the variances calculated to analyze the main effects
are like the between variances

Pharmacyst Page 41
Interaction Effect
The interaction effect is the effect that one factor has on the other factor. The degrees of freedom
here is the product of the two degrees of freedom for each factor.

Within Variation
The Within variation is the sum of squares within each treatment group. You have one less than
the sample size (remember all treatment groups must have the same sample size for a two-way
ANOVA) for each treatment group. The total number of treatment groups is the product of the
number of levels for each factor. The within variance is the within variation divided by its
degrees of freedom.
The within group is also called the error.

F-Tests
There is an F-test for each of the hypotheses, and the F-test is the mean square for each main
effect and the interaction effect divided by the within variance. The numerator degrees of
freedom come from each effect, and the denominator degrees of freedom is the degrees of
freedom for the within variance in each case.

Two-Way ANOVA Table

It is assumed that main effect A has a levels (and A = a-1 df), main effect B has b levels (and B =
b-1 df), n is the sample size of each treatment, and N = abn is the total sample size. Notice the
overall degrees of freedom is once again one less than the total sample size.
Source SS df MS F

Main Effect A given A, SS / df MS(A) / MS(W)

a-1

Main Effect B given B, SS / df MS(B) / MS(W)

b-1

Interaction Effect given A*B, SS / df MS(A*B) / MS(W)

(a-1)(b-1)

Within given N - ab, SS / df

ab(n-1)

Total sum of others N - 1,

abn - 1

Pharmacyst Page 42
Summary
The following results are calculated using the Quattro Pro spreadsheet. It provides the p-value
and the critical values are for alpha = 0.05.
Source of Variation SS df MS F P-value F-crit

Seed 512.8667 2 256.4333 28.283 0.000008 3.682

Fertilizer 449.4667 4 112.3667 12.393 0.000119 3.056

Interaction 143.1333 8 17.8917 1.973 0.122090 2.641

Within 136.0000 15 9.0667

Total 1241.4667 29

From the above results, we can see that the main effects are both significant, but the interaction
between them isn't. That is, the types of seed aren't all equal, and the types of fertilizer aren't all
equal, but the type of seed doesn't interact with the type of fertilizer.

ADDITIONAL:

Stats: Scheffe' and Tukey Tests

When the decision from the One-Way Analysis of Variance is to reject the null hypothesis, it
means that at least one of the means isn't the same as the other means. What we need is a way to
figure out where the differences lie, not just that there is a difference.
This is where the Scheffe' and Tukey tests come into play. They will help us analyze pairs of
means to see if there is a difference -- much like the difference of two means covered earlier.

Hypotheses

Both tests are set up to test if pairs of means are different. The formulas refer to
mean i and mean j. The values of i and j vary, and the total number of tests will
be equal to a combination of k objects, 2 at a time C(k,2), where k is the number
of samples.

Scheffé Test

Pharmacyst Page 43
The Scheffe' test is customarily used with unequal sample sizes, although it could be used with
equal sample sizes.
The critical value for the Scheffe' test is the degrees of freedom for the between variance times
the critical value for the one-way ANOVA. This simplifies to be:
CV = (k-1) F(k-1,N-k,alpha)
The test statistic is a little bit harder to compute. Pure
mathematicians will argue that this shouldn't be called F because
it doesn't have an F distribution (it's the degrees of freedom
times an F), but we'll live it with it.
Reject H0 if the test statistic is greater than the critical value.
Note, this is a right tail test. If there is no difference between the means, the numerator will be
close to zero, and so performing a left tail test wouldn't show anything.

Tukey Test

The Tukey test is only usable when the sample sizes are the same.
The Critical Value is looked up in a table. It is Table N in the Bluman text. There are actually
several different tables, one for each level of significance. The number of samples, k, is used as a
index along the top, and the degrees of freedom for the within group variance, v = N-k, are used
as an index along the left side.
The test statistic is found by dividing the difference between the means by
the square root of the ratio of the within group variation and the sample
size.
Reject the null hypothesis if the absolute value of the test statistic is
greater than the critical value (just like the linear correlation coefficient
critical values).

DEGREES OF FREEDOM:
In statistics, the number of degrees of freedom is the number of values in the final calculation of
a statistic that are free to vary.
Estimates of statistical parameters can be based upon different amounts of information or data.
The number of independent pieces of information that go into the estimate of a parameter is
called the degrees of freedom (df). In general, the degrees of freedom of an estimate of a
parameter is equal to the number of independent scores that go into the estimate minus the
number of parameters used as intermediate steps in the estimation of the parameter itself (which,
in sample variance, is one, since the sample mean is the only intermediate step).
Mathematically, degrees of freedom is the dimension of the domain of a random vector, or
essentially the number of 'free' components: how many components need to be known before the
vector is fully determined.
The term is most often used in the context of linear models (linear regression, analysis of
variance), where certain random vectors are constrained to lie in linear subspaces, and the
number of degrees of freedom is the dimension of the subspace. The degrees-of-freedom are also

Pharmacyst Page 44
commonly associated with the squared lengths (or "Sum of Squares") of such vectors, and the
parameters of chi-squared and other distributions that arise in associated statistical testing
problems.

ODDS RATIO:
The odds ratio is a measure of effect size, describing the strength of association or non-
independence between two binary data values. It is used as a descriptive statistic, and plays an
important role in logistic regression. Unlike other measures of association for paired binary data
such as the relative risk, the odds ratio treats the two variables being compared symmetrically,
and can be estimated using some types of non-random samples.
Definition in terms of group-wise odds
The odds ratio is the ratio of the odds of an event occurring in one group to the odds of it
occurring in another group. The term is also used to refer to sample-based estimates of this ratio.
These groups might be men and women, an experimental group and a control group, or any other
dichotomous classification. If the probabilities of the event in each of the groups are p1 (first
group) and p2 (second group), then the odds ratio is:

where qx = 1 − px. An odds ratio of 1 indicates that the condition or event under study is equally
likely to occur in both groups. An odds ratio greater than 1 indicates that the condition or event is
more likely to occur in the first group. And an odds ratio less than 1 indicates that the condition
or event is less likely to occur in the first group. The odds ratio must be nonnegative if it is
defined. It is undefined if p2q1 equals zero, i.e., if p2 equals zero or p1 equals one.
P VALUE:
Definition of a P value
In statistical significance testing, the p-value is the probability of obtaining a test statistic at least
as extreme as the one that was actually observed, assuming that the null hypothesis is true. One
often "rejects the null hypothesis" when the p-value is less than the significance level α (Greek
alpha), which is often 0.05 or 0.1. When the null hypothesis is rejected, the result is said to be
statistically significant.
A closely related concept is the E-value, which is the average number of times in multiple testing
that one expects to obtain a test statistic at least as extreme as the one that was actually observed,
assuming that the null hypothesis is true. The E-value is the product of the number of tests and
the p-value.
Although there is often confusion, the p-value is not the probability of the null hypothesis being
true, nor is the p-value the same as the Type I error rate, α.
Consider an experiment where you've measured values in two samples, and the means are
different. How sure are you that the population means are different as well? There are two
possibilities:

Pharmacyst Page 45
The populations have different means.
The populations have the same mean, and the difference you observed is a coincidence of
random sampling.
The P value is a probability, with a value ranging from zero to one. It is the answer to this
question: If the populations really have the same mean overall, what is the probability that
random sampling would lead to a difference between sample means as large (or larger) than you
observed?

How are P values calculated? There are many methods, and you'll need to read a statistics text to
learn about them. The choice of statistical tests depends on how you express the results of an
experiment (measurement, survival time, proportion, etc.), on whether the treatment groups are
paired, and on whether you are willing to assume that measured values follow a Gaussian bell-
shaped distribution.

Common misinterpretation of a P value

Many people misunderstand what question a P value answers.

If the P value is 0.03, that means that there is a 3% chance of observing a difference as large as
you observed even if the two population means are identical. It is tempting to conclude,
therefore, that there is a 97% chance that the difference you observed reflects a real difference
between populations and a 3% chance that the difference is due to chance. Wrong. What you can
say is that random sampling from identical populations would lead to a difference smaller than
you observed in 97% of experiments and larger than you observed in 3% of experiments.

You have to choose. Would you rather believe in a 3% coincidence? Or that the population
means are really different?

"Extremely significant" results

Intuitively, you probably think that P=0.0001 is more statistically significant than P=0.04. Using
strict definitions, this is not correct. Once you have set a threshold P value for statistical
significance, every result is either statistically significant or is not statistically significant. Some
statisticians feel very strongly about this.

Many scientists are not so rigid, and refer to results as being "very significant" or "extremely
significant" when the P value is tiny. Often, results are flagged with a single asterisk when the P
value is less than 0.05, with two asterisks when the P value is less than 0.01, and three asterisks
when the P value is less than 0.001. This is not a firm convention, so you need to check the
figure legends when you see asterisks to find the definitions the author used.

Pharmacyst Page 46
One- vs. two-tail P values

When comparing two groups, you must distinguish between one- and two-tail P values.

Start with the null hypothesis that the two populations really are the same and that the observed
discrepancy between sample means is due to chance.

The two-tail P value answers this question: Assuming the null hypothesis, what is the chance that
randomly selected samples would have means as far apart as observed in this experiment with
either group having the larger mean?
To interpret a one-tail P value, you must predict which group will have the larger mean before
collecting any data. The one-tail P value answers this question: Assuming the null hypothesis,
what is the chance that randomly selected samples would have means as far apart as observed in
this experiment with the specified group having the larger mean?
A one-tail P value is appropriate only when previous data, physical limitations or common sense
tell you that a difference, if any, can only go in one direction. The issue is not whether you
expect a difference to exist - that is what you are trying to find out with the experiment. The
issue is whether you should interpret increases and decreases the same.

You should only choose a one-tail P value when you believe the following:

Before collecting any data, you can predict which group will have the larger mean (if the means
are in fact different).
If the other group ends up with the larger mean, then you should be willing to attribute that
difference to chance, no matter how large the difference.
It is usually best to use a two-tail P value for these reasons:

The relationship between P values and confidence intervals is more clear with two-tail P values.
Some tests compare three or more groups, which makes the concept of tails inappropriate (more
precisely, the P values have many tails). A two-tail P value is more consistent with the P values
reported by these tests.
Choosing a one-tail P value can pose a dilemma. What would you do if you chose a one-tail P
value, but observed a large difference in the opposite direction to the experimental hypothesis?
To be rigorous, you should conclude that the difference is due to chance, and that the difference
is not statistically significant. But most people would be tempted to switch to a two-tail P value
or to reverse the direction of the experimental hypothesis. You avoid this situation by always
using two-tail P values.
Statistical hypothesis testing

Pharmacyst Page 47
The P value is a fraction. In many situations, the best thing to do is report that number to
summarize the results of a comparison. If you do this, you can totally avoid the term "statistically
significant", which is often misinterpreted.

In other situations, you'll want to make a decision based on a single comparison. In these
situations, follow the steps of statistical hypothesis testing.

Set a threshold P value before you do the experiment. Ideally, you should set this value based on
the relative consequences of missing a true difference or falsely finding a difference. In fact, the
threshold value (called alpha) is traditionally almost always set to 0.05.
Define the null hypothesis. If you are comparing two means, the null hypothesis is that the two
populations have the same mean.
Do the appropriate statistical test to compute the P value.
Compare the P value to the preset threshold value. If the P value is less than the threshold, state
that you "reject the null hypothesis" and that the difference is "statistically significant". If the P
value is greater than the threshold, state that you "do not reject the null hypothesis" and that the
difference is "not statistically significant".
Note that statisticians use the term hypothesis testing very differently than scientists.

Statistical significance
The term significant is seductive, and it is easy to misinterpret it. A result is said to be
statistically significant when the result would be surprising if the populations were really
identical. A result is said to be statistically significant when the P value is less than a preset
threshold value.

It is easy to read far too much into the word significant because the statistical use of the word has
a meaning entirely distinct from its usual meaning. Just because a difference is statistically
significant does not mean that it is important or interesting. And a result that is not statistically
significant (in the first experiment) may turn out to be very important.

If a result is statistically significant, there are two possible explanations:

The populations are identical, so there really is no difference. You happened to randomly obtain
larger values in one group and smaller values in the other, and the difference was large enough to
generate a P value less than the threshold you set. Finding a statistically significant result when
the populations are identical is called making a Type I error.
The populations really are different, so your conclusion is correct.
There are also two explanations for a result that is not statistically significant:

Pharmacyst Page 48
The populations are identical, so there really is no difference. Any difference you observed in the
experiment was a coincidence. Your conclusion of no significant difference is correct.
The populations really are different, but you missed the difference due to some combination of
small sample size, high variability and bad luck. The difference in your experiment was not large
enough to be statistically significant. Finding results that are not statistically significant when the
populations are different is called making a Type II error.
Confidence intervals

Statistical calculations produce two kinds of results that help you make inferences about the
populations from the samples. You've already learned about P values. The second kind of result
is a confidence interval.
95% confidence interval of a mean
Although the calculation is exact, the mean you calculate from a sample is only an estimate of
the population mean. How good is the estimate? It depends on how large your sample is and how
much the values differ from one another. Statistical calculations combine sample size and
variability to generate a confidence interval for the population mean. You can calculate intervals
for any desired degree of confidence, but 95% confidence intervals are used most commonly. If
you assume that your sample is randomly selected from some population, you can be 95% sure
that the confidence interval includes the population mean. More precisely, if you generate many
95% CI from many data sets, you expect the CI to include the true population mean in 95% of
the cases and not to include the true mean value in the other 5%. Since you don't know the
population mean, you'll never know for sure whether or not your confidence interval contains the
true mean.

Other situations

When comparing groups, calculate the 95% confidence interval for the difference between the
population means. Again interpretation is straightforward. If you accept the assumptions, there is
a 95% chance that the interval you calculate includes the true difference between population
means.

Methods exist to compute a 95% confidence interval for any calculated statistic, for example the
relative risk or the best-fit value in nonlinear regression. The interpretation is the same in all
cases. If you accept the assumptions of the test, you can be 95% sure that the interval contains
the true population value. Or more precisely, if you repeat the experiment many times, you
expect the 95% confidence interval will contain the true population value in 95% of the
experiments.

Why 95%?

Pharmacyst Page 49
There is nothing special about 95%. It is just convention that confidence intervals are usually
calculated for 95% confidence. In theory, confidence intervals can be computed for any degree
of confidence. If you want more confidence, the intervals will be wider. If you are willing to
accept less confidence, the intervals will be narrower.

In statistical significance testing, the p-value is the probability of obtaining a test statistic at least
as extreme as the one that was actually observed, assuming that the null hypothesis is true. One
often "rejects the null hypothesis" when the p-value is less than the significance level α (Greek
alpha), which is often 0.05 or 0.1. When the null hypothesis is rejected, the result is said to be
statistically significant.
A closely related concept is the E-value, which is the average number of times in multiple testing
that one expects to obtain a test statistic at least as extreme as the one that was actually observed,
assuming that the null hypothesis is true. The E-value is the product of the number of tests and
the p-value.
Although there is often confusion, the p-value is not the probability of the null hypothesis being
true, nor is the p-value the same as the Type I error rate, α.

Pharmacyst Page 50