1

Basic Statistics
■ Basic Statistics Overview, 1-2
■ Descriptive Statistics Available for Display or Storage, 1-4
■ Display Descriptive Statistics, 1-6
■ Store Descriptive Statistics, 1-9
■ One-Sample Z-Confidence Interval and Test, 1-10
■ One-Sample t-Confidence Interval and Test, 1-14
■ Two-Sample t-Confidence Interval and Test, 1-17
■ Paired t-Confidence Interval and Test, 1-21
■ Confidence Interval and Test of a Proportion, 1-24
■ Confidence Interval and Test of Two Proportions, 1-28
■ Correlation, 1-32
■ Covariance, 1-36
■ Normality Test, 1-37

To Table of Contents To Index

MINITAB User’s Guide 2 1-1

Chapter 1 Basic Statistics Overview

Basic Statistics Overview

Use MINITAB’s basic statistics capabilities for calculating basic statistics and for simple
estimation and hypothesis testing with one or two samples. The basic statistics
capabilities include procedures for
■ calculating or storing descriptive statistics
■ hypothesis tests and confidence intervals of the mean or difference in means
■ hypothesis tests and confidence intervals for a proportion or the difference in
proportions
■ measuring association
■ testing for normality of a distribution

Calculating and storing descriptive statistics

■ Display Descriptive Statistics produces descriptive statistics for each column or
subset within a column. You can print the statistics in the Session window and/or
display them in a graph.
■ Store Descriptive Statistics stores descriptive statistics for each column or subset
within a column.
For a list of descriptive statistics available for display or storage see page 1-4. To
calculate descriptive statistics individually and store them as constants, see the
Calculations chapter in MINITAB User’s Guide 1.

Confidence intervals and hypothesis tests of means

The four procedures for hypothesis tests and confidence intervals for population means
or the difference between means are based upon the distribution of the mean following
a normal distribution. According to the Central Limit Theorem, the normal
distribution becomes an increasingly better approximation for the distribution of the
mean of a sample drawn from any distribution as the sample size increases.
■ 1-Sample Z computes a confidence interval or performs a hypothesis test of the
mean when the population standard deviation, σ, is known. This procedure is based
upon the normal distribution, so for small samples, this procedure works best if your
data were drawn from a normal distribution or one that is close to normal. From the
Central Limit Theorem, you may use this procedure if you have a large sample,
substituting the sample standard deviation for σ. A common rule of thumb is to
consider samples of size 30 or higher to be large samples. Many analysts choose the
t-procedure over the Z-procedure whenever σ is unknown.

1-2 MINITAB User’s Guide 2

Basic Statistics Overview Basic Statistics

■ 1-Sample t computes a confidence interval or performs a hypothesis test of the

mean when σ is unknown. This procedure is based upon the t-distribution, which is
derived from a normal distribution with unknown σ. For small samples, this
procedure works best if your data were drawn from a distribution that is normal or
close to normal. This procedure is more conservative than the Z-procedure and
should always be chosen over the Z-procedure with small sample sizes and an
unknown σ. Many analysts choose the t-procedure over the Z-procedure anytime σ
is unknown. According to the Central Limit Theorem, you can have increasing
confidence in the results of this procedure as sample size increases, because the
distribution of the sample mean becomes more like a normal distribution.
■ 2-Sample t computes a confidence interval and performs a hypothesis test of the
difference between two population means when σ’s are unknown and samples are
drawn independently from each other. This procedure is based upon the
t-distribution, and for small samples it works best if data were drawn from
distributions that are normal or close to normal. You can have increasing confidence
of the results as the sample sizes increases.
■ Paired t computes a confidence interval and performs a hypothesis test of the
difference between two population means when observations. When data are
paired, as with before-and-after measurements, the paired t-procedure results in a
smaller variance and greater power of detecting differences than would the above
2-sample t-procedure, which assumes that the samples were independently drawn.

Confidence intervals and hypothesis tests of proportions

■ 1 Proportion computes a confidence interval and performs a hypothesis test of a
population proportion.
■ 2 Proportions computes a confidence interval and performs a hypothesis test of the
difference between two population proportions.

Measures of association
■ Correlation calculates the Pearson product moment coefficient of correlation (also
called the correlation coefficient or correlation) for pairs of variables. The
correlation coefficient is a measure of the degree of linear relationship between two
variables. You can obtain a p-value for testing if there is sufficient evidence that the
correlation coefficient is not zero.
By using a combination of MINITAB commands, you can also compute a partial
correlation coefficient which is a case of correlation. A partial correlation coefficient
is the correlation coefficient between two variables while adjusting for the effects of
other variables.

MINITAB User’s Guide 2 1-3

Chapter 1 Descriptive Statistics Available for Display or Storage

■ Covariance calculates the covariance for pairs of variables. The covariance is a

measure of the relationship between two variables but it has not been standardized,
as is done with the correlation coefficient, by dividing by the standard deviation of
both variables.

Distribution test
■ Normality Test generates a normal probability plot and performs a hypothesis test
to examine whether or not the observations follow a normal distribution. Some
statistical procedures, such as a z- or t-test, assume that the sample was drawn from a
normal distribution. Use this procedure to test the normality assumption.

Descriptive Statistics Available for Display or

Storage
The following table shows the descriptive statistics that you can display in the Session
window, in a graphical summary, or that you can store. When you display statistics, you
get all of the indicated statistics (see Display Descriptive Statistics on page 1-6); when
you store statistics, you can choose which ones to store (see Store Descriptive Statistics
on page 1-9).

Session Graphical
Statistic window summary Store
Number of nonmissing values ✗ ✗ ✗

Number of missing values ✗ ✗

Total number ✗

Cumulative number ✗

Percent ✗

Cumulative percent ✗

Mean ✗ ✗ ✗

Trimmed mean ✗

Confidence interval for µ ✗

Standard error of mean ✗ ✗

Standard deviation ✗ ✗ ✗

1-4 MINITAB User’s Guide 2

Descriptive Statistics Available for Display or Storage Basic Statistics

Session Graphical
Statistic window summary Store
Confidence interval for σ ✗

Variance ✗ ✗

Sum ✗

Minimum ✗ ✗ ✗

Maximum ✗ ✗ ✗

Range ✗

Median ✗ ✗ ✗

Confidence interval for median ✗

First and third quartiles ✗ ✗ ✗

Interquartile range ✗

Sums of squares ✗

Skewness ✗ ✗

Kurtosis ✗ ✗

MSSD ✗

Normality test statistic, p-value ✗

Calculations
Trimmed Mean. To calculate the trimmed mean, MINITAB removes the smallest 5%
and the largest 5% of the values (rounded to the nearest integer), and then averages the
remaining data.
Standard Error of Mean. Calculated by StDev ⁄ N .
Standard Deviation. If the column contains x1, x2, …, xn, with mean x , then

2
standard deviation = Σ( x – x ) ⁄ (n – 1 )
Confidence Interval for σ. The confidence interval for σ is

2 2
( n – 1 )s ( n – 1 )s
------------------------------- to ------------------------
2 2
χ n – 1, 1 – α ⁄ 2 χ n – 1, α ⁄ 2

MINITAB User’s Guide 2 1-5

Chapter 1 Display Descriptive Statistics

2
Variance. The standard deviation squared or Σ ( x – x ) ⁄ ( n – 1 ) .
Median. If sample size is odd, the median is the (n+1) / 2th ordered value. If sample
size is even, the median is the mean of the two middle ordered values.
Confidence Interval for Median. Uses one-sample sign confidence interval described
on page 5-3.
Quartiles. To calculate quartiles, MINITAB orders the data from smallest to largest. The
first quartile (Q1) is the observation at position (n + 1) / 4, and the third quartile (Q3) is
the observation at position 3(n + 1) / 4, where n is the number of observations. If the
position is not an integer, interpolation is used.
Sums of Squares. This is the uncorrected sum of squares, or the sum of squared data
values.
Skewness. This is a measure of distribution asymmetry or the tendency of one tail to be
heavier than the other. A negative value indicates skewness to the left and a positive
values indicates skewness to the right, though a value of zero does not necessarily
indicate symmetry. Skewness is calculated as
3 3
n ⁄ ( n – 1 ) ( n–2 ) Σ ( x – x ) ⁄ n
[Link] is a measure of how different a distribution is from the normal
distribution. A negative value typically indicates a distribution with heavier tails than
the normal. A positive value typically indicates a distribution with less heavy tails than
the normal. Kurtosis is calculated as
4 4 2
n ( n + 1 ) ⁄ ( n – 1 ) ( n – 2 ) ( n – 3 )Σ ( x – x ) ⁄ s – 3 ( n – 1 ) ⁄ ( n – 2 ) ( n – 3 )
MSSD. This is half the Mean of Successive Squared Differences. For example, if the
data are 1, 2, 4, 10, successive differences are 1, 2, 6, and the MSSD is
(mean of 12, 22, 62) / 2, or 6.833

Display Descriptive Statistics

Use Display Descriptive Statistics to produce statistics for each column or for subsets
within a column. You can display these statistics in the Session window and optionally
in a graph. See Descriptive Statistics Available for Display or Storage on page 1-4.

Data
The data columns must be numeric. The optional grouping column (also called a By
column) can be numeric, text, or date/time and must be the same length as the data
columns. If you wish to change the order in which text categories are processed from

1-6 MINITAB User’s Guide 2

Display Descriptive Statistics Basic Statistics

their default alphabetical order, you can define your own order. See Ordering Text
Categories in the Manipulating Data chapter in MINITAB User’s Guide 1.
MINITAB automatically omits missing data from the calculations.

h To calculate descriptive statistics

1 Choose Stat ➤ Basic Statistics ➤ Display Descriptive Statistics.

2 In Variables, enter the column(s) containing the data you want to describe.

3 If you like, use one or more of the options listed below, then click OK.

Options

Descriptive Statistics dialog box

■ display separate statistics for each unique value in a By column.

Graphs subdialog box

■ generate a histogram, a histogram with a normal curve, a dotplot, or a boxplot of the
data in separate Graph windows.
■ display statistics in a single graphical summary. You can specify the confidence level
for the displayed confidence interval. The default level is 95%.
See Descriptive statistics graphs below for details.

Descriptive statistics graphs

You can display your data in a histogram, a histogram with normal curve, a dotplot, or a
boxplot, or display a graphical summary. The displayed statistics are listed in
Descriptive Statistics Available for Display or Storage on page 1-4.

MINITAB User’s Guide 2 1-7

Chapter 1 Display Descriptive Statistics

The graphical summary includes a table of descriptive statistics, a histogram with

normal curve, a boxplot, a confidence interval for the population mean, µ, and a
confidence interval for the population median. MINITAB can display a maximum of 15
graphs at a time. Therefore, the graphical summary will not work when there are more
than 15 columns, 15 distinct levels or groups in a By column, or the combination of
columns and By levels is more than 15.
There is no restriction on the number of columns or levels when producing output in
the Session window.

Tip If you exceed the maximum number of graphs because of the number of levels of your By
variable, you can decrease the number of graphs by unstacking your data and displaying
descriptive statistics for data subsets (see the Manipulating Data chapter in MINITAB User’s
Guide 1).

e Example of displaying descriptive statistics

You want to examine characteristic of the height (in inches) of male (Sex = 1) and
female (Sex = 2) students who participated in the pulse study. You choose to display
descriptive statistics with the option of a boxplot of the data.
1 Open the file [Link].

2 Choose Stat ➤ Basic Statistics ➤ Display Descriptive Statistics.

3 In Variables, enter Height. Check By variable and enter Sex in the text box.

4 Click Graphs. Check Boxplot of data. Click OK twice.

Session Descriptive Statistics

window
output Variable Sex N Mean Median TrMean StDev
Height 1 57 70.754 71.000 70.784 2.583
2 35 65.400 65.500 65.395 2.563

Variable Sex SE Mean Minimum Maximum Q1 Q3

Height 1 0.342 66.000 75.000 69.000 73.000
2 0.433 61.000 70.000 63.000 68.000

Graph
window
output

1-8 MINITAB User’s Guide 2

Store Descriptive Statistics Basic Statistics

Interpreting the results

The means shown in the Session window and the boxplots indicate that males are
approximately 5.3 inches taller than females, and the spread of the data is about the
same.

Store Descriptive Statistics

You can store descriptive statistics for each column or for subsets within a column. See
Descriptive Statistics Available for Display or Storage on page 1-4.

Data
The data columns must be numeric. The optional grouping column (also called a By
column) can be numeric, text, or date/time and must be the same length as the data
columns. If you wish to change the order in which text categories are processed from
their default alphabetical order, you can define your own order. See Ordering Text
Categories in the Manipulating Data chapter in MINITAB User’s Guide 1.
MINITAB automatically omits missing data from the calculations.

h To store descriptive statistics

1 Choose Stat ➤ Basic Statistics ➤ Store Descriptive Statistics.

2 In Variables, enter the column(s) containing the data you want to describe.

3 If you like, use one or more of the options listed below, then click OK.

MINITAB User’s Guide 2 1-9

Chapter 1 One-Sample Z-Confidence Interval and Test

Options

Descriptive Statistics dialog box

■ calculate statistics corresponding to values in one or more By columns.

Statistics subdialog box

■ select the statistics that you wish to store. The defaults are sample mean and sample
size (nonmissing).

Options subdialog box

When you use a By variable, you can also
■ store statistics for empty cells of a cross-classification of the By variables. This is the
default.
■ include missing data as a valid By variable classification.
■ store the distinct values of the By variables. This is the default.

Naming stored columns

MINITAB automatically names the storage columns with the name of the stored statistic
and a sequential integer starting at 1. For example, suppose you enter two columns in
Variables and choose to store the default mean and sample size. MINITAB will name
the storage columns Mean1 and N1 for the first variable and Mean2 and N2 for the
second variable. If you use two By variables, MINITAB will store the distinct levels
(subscripts) of the By variables in columns named ByVar1 and ByVar2, with the
appended integer cycling as with the stored statistics.
If you erase the storage columns or rename them, the integers will start over at 1. If you
store statistics for many columns, you may want to rename the corresponding stored
columns so that you can keep track of their origin.

One-Sample Z-Confidence Interval and Test

Use 1-Sample Z to compute a confidence interval or perform a hypothesis test of the
mean when σ is known, for sample data in one or more columns. For a two-tailed
one-sample Z
H0: µ = hypothesized µ versus H1: µ ≠ hypothesized µ
where µ is the population mean.

1-10 MINITAB User’s Guide 2

One-Sample Z-Confidence Interval and Test Basic Statistics

Data
Enter each sample in a single numeric column. You can generate a hypothesis test or
confidence interval for more than one column at a time.
MINITAB automatically omits missing data from the calculations for that column in
which it is missing.

h To do a Z-confidence interval and test of the mean

1 Choose Stat ➤ Basic Statistics ➤ 1-Sample Z.

2 In Variables, enter the column(s) containing the samples.

3 Choose one of the following:

■ to calculate a confidence interval for the mean, choose Confidence interval
■ to perform a hypothesis test, choose Test mean
4 In Sigma, enter a value for σ.

5 If you like, use one or more of the options listed below, then click OK.

Options

1-Sample Z dialog box

■ specify a confidence level for the confidence interval. The default is 95%.
■ specify a null hypothesis test value. The default is 0.
■ define the alternative hypothesis by choosing less than (lower-tailed), not equal
(two-tailed), or greater than (upper-tailed). The default is a two-tailed test.

MINITAB User’s Guide 2 1-11

Chapter 1 One-Sample Z-Confidence Interval and Test

Graphs subdialog box

■ display a histogram, dotplot, and boxplot for each column. The graphs show the
sample mean and a confidence interval for the mean. When you do a hypothesis
test, the graphs also show the null hypothesis test value.

Method

Confidence interval
The confidence interval is calculated as

x – z α ⁄ 2 ( σ ⁄ n ) to x + z α ⁄ 2 ( σ ⁄ n )

where x is the mean of the data, σ is the population standard deviation, n is the sample
size, and zα/2 is the value from the normal table where α is 1 − confidence level / 100.
You can specify a confidence level by entering any number between 1 and 100 in
Level. The confidence level is 95% by default.

Hypothesis test
MINITAB calculates the test statistic by
x–µ
Z = --------------
σ⁄ n
where x is the mean of the data, µ is the hypothesized population mean, σ is the
population standard deviation, and n is the sample size.
When µ is not specified in Test mean, µ = 0 is used. MINITAB performs a two-tailed test
unless you specify a one-tailed test.

e Example of one-sample Z-confidence interval and test

Measurements were made on nine widgets. You know that the distribution of
measurements has historically been close to normal with σ = 0.2. Since you know σ,
and you wish to test if the population mean is 5 and obtain a 90% confidence interval
for the mean, you use the Z-procedure.

Step 1: Perform the Z-test with a dotplot of the data

1 Open the file EXH_STAT.MTW.

2 Choose Stat ➤ Basic Statistics ➤ 1-Sample Z.

3 In Variables, enter Values.

4 Choose Test mean and enter 5.

1-12 MINITAB User’s Guide 2

One-Sample Z-Confidence Interval and Test Basic Statistics

5 In Sigma, enter 0.2.

6 Click Graphs. Check Dotplot of data. Click OK twice.

Step 2: Obtain the Z-confidence interval with a histogram of the data

1 Repeat 1-3 above.

2 Choose Confidence interval, and in Level, enter 90.

3 In Sigma, enter 0.2.

4 Click Graphs.

5 If Dotplot of data is checked, uncheck it. Check Histogram of the data. Click OK
twice.

Session Z-Test
window
output Test of mu = 5.0000 vs mu not = 5.0000
The assumed sigma = 0.200

Variable N Mean StDev SE Mean Z P

Values 9 4.7889 0.2472 0.0667 -3.17 0.0016

Graph
window
output

Session Z Confidence Intervals

window
output The assumed sigma = 0.200

Variable N Mean StDev SE Mean 90.0 % CI

Values 9 4.7889 0.2472 0.0667 ( 4.6792, 4.8986)

MINITAB User’s Guide 2 1-13

Chapter 1 One-Sample t-Confidence Interval and Test

Graph
window
output

Interpreting the results

The test statistic, Z, for testing if the population mean = 5 is −3.17. The p-value of the
test, or the probability of obtaining a more extreme value of the test statistic by chance
if the null hypothesis was true, is 0.0016. This is called the attained significance level,
p-value, or attained α of the test. Since the p-value of 0.0016 is smaller than likely α
levels, there is significant evidence that µ is not equal to 5, so we reject H0 in favor of µ
not being 5.
A hypothesis test at α = 0.05 could also be performed by viewing the dotplot and noting
that the hypothesized value falls outside the confidence interval.
A 90% confidence interval for the population mean, µ, is (4.6792, 4.8986). The
histogram also shows this interval.

One-Sample t-Confidence Interval and Test

Use 1-Sample t to compute a confidence interval and perform a hypothesis test of the
mean when the population standard deviation, σ, is unknown. For a two-tailed
one-sample t,
H0: µ = hypothesized µ versus H1: µ ≠ hypothesized µ
where µ is the population mean.

Data
Enter each sample in a single numeric column. You can generate a hypothesis test or
confidence interval for more than one column at a time.
MINITAB automatically omits missing data from the calculations.

1-14 MINITAB User’s Guide 2

One-Sample t-Confidence Interval and Test Basic Statistics

h To compute a t-confidence interval and test of the mean

1 Choose Stat ➤ Basic Statistics ➤ 1-Sample t.

2 In Variables, enter the column(s) containing the samples.

3 Do one of the following:

■ to calculate a confidence interval for the mean, choose Confidence interval
■ to perform a hypothesis test, choose Test mean
4 If you like, use one or more of the options listed below, then click OK.

Options

1-Sample t dialog box

■ specify a confidence level for the confidence interval. The default is 95%.
■ specify a null hypothesis test value. The default is 0.
■ define the alternative hypothesis by choosing less than (lower-tailed), not equal
(two-tailed), or greater than (upper-tailed). The default is a two-tailed test.

Graphs subdialog box

■ display a histogram, dotplot, and boxplot for each column. The graphs show the
sample mean and a confidence interval for the mean and additionally, the null
hypothesis test value when you do a hypothesis test.

MINITAB User’s Guide 2 1-15

Chapter 1 One-Sample t-Confidence Interval and Test

Method

Confidence interval
The confidence interval is calculated as

x – t α ⁄ 2 ( s ⁄ n ) to x + t α ⁄ 2 ( s ⁄ n )

where x is the mean of the data, s is the sample standard deviation, n is the sample
size, and tα/2 is the value from a t-distribution table where α is 1 − confidence level /
100 and degrees of freedom are (n − 1).
You can specify a confidence level by entering any number between 1 and 100 in
Level. The confidence level is 95% by default.

Hypothesis test
MINITAB calculates the test statistic by

x–µ
t = -------------
s⁄ n
where x is the mean of the data, µ is the estimated population mean, s is the sample
standard deviation, and n is the sample size.
When µ is not specified in Test mean, µ = 0 is used. MINITAB performs a two-tailed test
unless you specify a one-tailed test.

e Example of a one-sample t-confidence interval and test

Measurements were made on nine widgets. You know that the distribution of
measurements has historically been close to normal, but suppose that you do not know
σ. To test if the population mean is 5 and to obtain a 90% confidence interval for the
mean, you use a t-procedure.

Step 1: Perform the t-test

1 Open the file EXH_STAT.MTW.

2 Choose Stat ➤ Basic Statistics ➤ 1-Sample t.

3 In Variables, enter Values.

4 Choose Test mean and enter 5. Click OK.

Step 2: Obtain the t-confidence interval

1 Repeat 1–3 above.

2 Choose Confidence interval, and in Level, enter 90. Click OK.

1-16 MINITAB User’s Guide 2

Two-Sample t-Confidence Interval and Test Basic Statistics

Session T-Test of the Mean

window
output Test of mu = 5.0000 vs mu not = 5.0000

Variable N Mean StDev SE Mean T P-Value

Values 9 4.7889 0.2472 0.0824 -2.56 0.033

Session T Confidence Intervals

window
output Variable N Mean StDev SE Mean 90.0 % C.I.
Values 9 4.7889 0.2472 0.0824 ( 4.6356, 4.9422)

Interpreting the results

The test statistic, T, for H0: µ = 5 is calculated as −2.56.
The p-value of this test, or the probability of obtaining more extreme value of the test
statistic by chance if the null hypothesis was true, is 0.033. This is called the attained
significance level, or p-value. Therefore, reject H0 if your acceptable α level is greater
than the p-value, or 0.033.
A 90% confidence interval for the population mean, µ, is (4.6356, 4.9422). This
interval is slightly wider than the corresponding Z-interval shown in Example of
one-sample Z-confidence interval and test on page 1-12.

Two-Sample t-Confidence Interval and Test

Use 2-Sample t to perform a hypothesis test and compute a confidence interval of the
difference between two population means when the population standard deviations,
σ’s, are unknown. For a two-tailed two-sample t
H0: µ1 = µ2 versus H1: µ1 ≠ µ2
where µ is the population mean.

Data
Data can be entered in one of two ways:
■ both samples in a single numeric column with another grouping column (called
subscripts) to identify the population. The grouping column may be numeric, text,
or date/time.
■ each sample in a separate numeric column.
The sample sizes do not need to be equal. MINITAB automatically omits missing data
from the calculations.

MINITAB User’s Guide 2 1-17

Chapter 1 Two-Sample t-Confidence Interval and Test

h To do a two-sample test and confidence interval

1 Choose Stat ➤ Basic Statistics ➤ 2-Sample t.

2 Choose one of the following:

■ If your data are stacked in a single column:
– choose Samples in one column
– in Samples, enter the column containing the numeric data
– in Subscripts, enter the column containing the group or population codes
■ If your data are unstacked, that is each sample is in a separate column:
– choose Samples in different columns
– in First, enter the column containing the first sample
– in Second, enter the column containing the other sample
3 If you like, use one or more of the options listed below, and click OK.

Options

2-Sample t dialog box

■ define the alternative hypothesis by choosing less than (lower-tailed), not equal
(two-tailed), or greater than (upper-tailed). The default is a two-tailed test.
■ specify a confidence level for the confidence interval. The default is 95%.
■ assume that the populations have equal variances. The default is to assume unequal
variances. See Equal or unequal variances on page 1-19.

Graphs subdialog box

■ display a dotplot or boxplot of each sample in the same [Link] graphs also
display the sample means.

1-18 MINITAB User’s Guide 2

Two-Sample t-Confidence Interval and Test Basic Statistics

Equal or unequal variances

If you check Assume equal variances, the sample standard deviations are pooled to
obtain a single estimate of σ. See Standard deviations under Method below for
calculations.
The two-sample t-test with a pooled variances is slightly more powerful than the
two-sample t-test with unequal variances, but serious error can result if the variances
are not equal. Therefore, the pooled variance estimate should not be used in many
cases. Use Variance Test on page 3-48 to test the equal variance assumption.

Method

Confidence interval
The confidence interval is calculated as
( x 1 – x 2 ) – t α ⁄ 2 s to ( x 1 – x 2 ) + t α ⁄ 2 s

where tα/2 is the value from a t-distribution table where α is 1 - confidence level/100.
The sample standard deviation, s, of x 1 − x 2 and the degrees of freedom depend upon
the variance assumption.
You can specify a confidence level of any number between 1 and 100 in Confidence
level. The confidence level is 95% by default.

Hypothesis test
MINITAB calculates the test statistic, t, by
t = ( x 1 − x 2 )/s
The sample standard deviation, s, of x 1 − x 2 depends upon the variance assumption.

Standard deviations
When you assume unequal variances, the sample standard deviation of x 1 − x 2 is
2 2
s1 s2
s = ------
- + ------
-
n 1 n2

The test statistic degrees of freedom are

2
( VAR 1 + VAR 2 )
df = -------------------------------------------------------------------------------------------------------------
-
2 2
[ ( VAR 1 ) ⁄ ( n 1 – 1 ) ] + [ ( VAR 2 ) ⁄ ( n 2 – 1 ) ]

MINITAB User’s Guide 2 1-19

Chapter 1 Two-Sample t-Confidence Interval and Test

where VAR1 = s12/n1, and VAR2 = s22/n2. MINITAB truncates the degrees of freedom to
an integer, if necessary. This is a more conservative approach than rounding.
When you assume equal variances, the pooled sample standard deviation of x 1 − x 2 is

2 2
( n 1 – 1 )s 1 + ( n 2 – 1 )s 2
sp = ------------------------------------------------------------
-
n1 + n2 – 2

The test statistic degrees of freedom are (n1 + n2 − 2).

e Example of a two-sample t-confidence interval and test

A study was performed in order to evaluate the effectiveness of two devices for
improving the efficiency of gas home-heating systems. Energy consumption in houses
was measured after one of the two devices was installed. The two devices were an
electric vent damper (Damper = 1) and a thermally activated vent damper (Damper =
2). The energy consumption data ([Link]) is stacked in one column with a grouping
column (Damper) containing identifiers or subscripts to denote the population.
Previously, you performed a variance test and found no evidence for variances being
unequal. Now you want to compare the effectiveness of these two devices by
determining whether or not there is any evidence that the difference between the
devices is different from zero.
1 Open the file [Link].

2 Choose Stat ➤ Basic Statistics ➤ 2-Sample T.

3 Choose Samples in one column.

4 In Samples, enter '[Link]'.

5 In Subscripts, enter Damper.

6 Check Assume equal variances. Click OK.

Session Two Sample T-Test and Confidence Interval

window
output Two sample T for [Link]
Damper N Mean StDev SE Mean
1 40 9.91 3.02 0.48
2 50 10.14 2.77 0.39

95% CI for mu (1) - mu (2): ( -1.45, 0.98)

T-Test mu (1) = mu (2) (vs not =): T= -0.38 P=0.70 DF= 88
Both use Pooled StDev = 2.88

Interpreting the result

MINITAB displays a table of the sample sizes, sample means, standard deviations, and
standard errors for the two samples.

1-20 MINITAB User’s Guide 2

Paired t-Confidence Interval and Test Basic Statistics

Since we previously found no evidence for variances being unequal, we chose to use
the pooled standard deviation by choosing Assume equal variances. The pooled
standard deviation, 2.88, is used to calculate the test statistic and the confidence
intervals.
A second table gives a confidence interval for the difference in population means. For
this example, a 95% confidence interval is (−1.45, 0.98). Next is the hypothesis test
result. The test statistic is −0.38, with p-value of 0.70, and degrees of freedom 88.
Since the p-value is greater than any likely α levels, there is no evidence for a difference
in energy use when using an electric vent damper versus a thermally activated vent
damper.

Paired t-Confidence Interval and Test

Use the Paired t command to compute a confidence interval and perform a hypothesis
test of the difference between population means when observations are paired. A paired
t-procedure matches responses that are dependent or related in a pairwise manner.
This matching allows you to account for variability between the pairs usually resulting
in a smaller error term, thus increasing the sensitivity of the hypothesis test or
confidence interval.
Typical examples of paired data include measurements on twins or before-and-after
measurements. For a paired t-test:
H0: µ1 = µ2 versus H1: µ1 ≠ µ2 where µ is the population mean.
When the samples are drawn independently from two populations, use the two-sample
t-procedure (page 1-17).

Data
The data from each sample must be in separate numeric columns of equal length.
Each row contains the paired measurements for an observation. If either measurement
of a row is missing, MINITAB automatically omits that row from the calculations.

MINITAB User’s Guide 2 1-21

Chapter 1 Paired t-Confidence Interval and Test

h To compute a paired t-confidence interval and test

1 Choose Stat ➤ Basic Statistics ➤ Paired t.

2 In First Sample, enter the column containing the first sample.

3 In Second Sample, enter the column containing the second sample.

4 If you like, use one or more of the options listed below, and click OK.

Options

Graphs subdialog box

■ display a histogram, dotplot, and boxplot of the paired differences.

Options subdialog box

■ specify a confidence level for the confidence interval. The default is 95%.
■ specify a null hypothesis test value. The default is 0.
■ define the alternative hypothesis by choosing less than (lower-tailed), not equal
(two-tailed), or greater than (upper-tailed). The default is a two-tailed test.

Method

Confidence interval
For a two-tailed test, the confidence interval is calculated as

d – t α ⁄ 2 ( s d ⁄ n ) to d + t α ⁄ 2 ( s d ⁄ n )

1-22 MINITAB User’s Guide 2

Paired t-Confidence Interval and Test Basic Statistics

where:

d = Σd ⁄ n , where d = x1 − x2 and x1 and x2 are observations from

populations 1 and 2, respectively
tα/2 is the value from a t-distribution where α is 1 − confidence level / 100
sd = the standard deviation of the differences
n = number of pairs of values

The standard deviation of the differences is calculated by:

2
sd = ∑ (d – d) -
-------------------------
(n – 1)
You can specify a confidence level of any number between 1 and 100. The confidence
level is 95% by default.

Hypothesis test
MINITAB calculates the test statistic, t, by:
d – µd
t = --------------------
- where µd is the hypothesized difference.
( sd ⁄ n )

When µd is not specified in Test mean, µd = 0 is used. MINITAB performs a two-tailed

test unless you specify a one-tailed test.

e Example of a confidence interval and test for paired data

A shoe company wants to compare two materials, A and B, for use on the soles of boys’
shoes. In this example, each of ten boys in a study wore a special pair of shoes with the
sole one shoe made from Material A and the sole on the other shoe from Material B.
The sole types were randomly assigned to account for systematic differences between
the left and right foot wear. After three months, the shoes are measured for wear.
For this data, you would use a paired design rather than an unpaired design. A paired
t-procedure would probably have a smaller error term than the corresponding unpaired
procedure because it removes variability that is due to differences between the pairs.
For example, one boy may live in the city and walk on pavement most of the day, while
another boy may live in the country and spend much of his day on unpaved surfaces.
1 Open the file EXH_STAT.MTW.

2 Choose Stat ➤ Basic Statistics ➤ Paired t.

3 In First Sample, enter Mat-A. In Second Sample, enter Mat-B. Click OK.

MINITAB User’s Guide 2 1-23

Chapter 1 Confidence Interval and Test of a Proportion

Session Paired T-Test and Confidence Interval

window
output Paired T for Mat-A - Mat-B

N Mean StDev SE Mean

Mat-A 10 10.630 2.451 0.775
Mat-B 10 11.040 2.518 0.796
Difference 10 -0.410 0.387 0.122

95% CI for mean difference: (-0.687, -0.133)

T-Test of mean difference = 0 (vs not = 0): T = -3.35 P-Value = 0.009

Interpreting the results

The confidence interval for the difference between the two materials does not include
zero, which suggests a difference between them. The small p-value (p = 0.009) further
suggests that the data are inconsistent with H0: µ1 = µ2, that is, the two materials do not
perform equally. Specifically, Material B ( X = 11.04) performed better than Material A
( X = 10.63) in terms of wear over the three month test period.
If you compare the results from the paired procedure with those from the unpaired
procedure (Stat ➤ Basic Statistics ➤ 1-Sample t). For the paired procedure, you would
conclude that the data are inconsistent with H0 (t = −3.35; p = 0.009). For the unpaired
procedure, you would conclude that the data are consistent with H0 (t = −0.37; p =
0.72). (These results are not shown.) That is, you would not have been able to detect
any difference in wear between the two materials.

Confidence Interval and Test of a Proportion

Use the 1 Proportion command to compute a confidence interval and perform a
hypothesis test of the proportion. For example, an automotive parts manufacturer
claims that his spark plugs are less than 2% defective. You could take a random sample
of spark plugs and determine whether or not the actual proportion defective is
consistent with the claim. For a two-tailed test of a proportion:
H0: p = p0 versus H1: p ≠ p0 where p is the population proportion and p0 is the
hypothesized value.
To compare two proportions, use Stat ➤ Basic Statistics ➤ 2 Proportions described on
page 1-28.

Data
You can have data in two forms: raw or summarized.

1-24 MINITAB User’s Guide 2

Confidence Interval and Test of a Proportion Basic Statistics

Raw data
Enter each sample in a numeric, text, or date/time column in your worksheet.
Columns must be all of the same type. Each column contains both the success and
failure data for that sample. Successes and failures are determined by numeric or
alphabetical order. MINITAB defines the lowest value as the failure; the highest value as
the success. For example:
■ for the numeric column entries of “20” and “40,” observations of 20 are considered
failures; observations of 40 are considered successes.
■ for the text column entries of “alpha” and “omega,” observations of alpha are
considered failures; observations of omega are considered successes. If the data
entries are “red” and “yellow,” observations of red are considered failures;
observations of yellow are considered successes.
You can reverse the definition of success and failure in a text column by applying a
value order—see Ordering Text Categories in the Manipulating Data chapter of
MINITAB User’s Guide 1.
With raw data, you can generate a hypothesis test or confidence interval for more than
one column at a time. When you enter more than one column, MINITAB performs a
separate analysis for each column.
MINITAB omits missing data from the calculations.

Summarized data
Enter the number of trials and one or more values for the number of successes directly
in the 1 Proportion dialog box. When you enter more than one success value, MINITAB
performs a separate analysis for each one.

h To calculate a confidence interval and test of a proportion

1 Choose Stat ➤ Basic Statistics ➤ 1 Proportion.

MINITAB User’s Guide 2 1-25

Chapter 1 Confidence Interval and Test of a Proportion

2 Do one of the following:

■ If you have raw data:
1 Choose Samples in columns, and enter the columns containing the raw data.

■ If you have summarized data:

1 Choose Summarized data.
2 In Number of trials, enter a whole number.
3 In Number of successes, enter one or more whole numbers.

4 If you like, use one or more of the options listed below, and click OK.

Options

Options subdialog box

■ specify a confidence level for the confidence interval. The default is 95%.
■ specify a null hypothesis test value. The default is 0.5.
■ define the alternative hypothesis by choosing less than (lower-tailed), not equal
(two-tailed), or greater than (upper-tailed). The default is a two-tailed test.
■ use a normal approximation rather than the exact test for both the hypothesis test
and confidence interval. See Method below.

Method

Confidence interval
By default, MINITAB uses an exact method [4] to calculate the confidence interval
limits (pL, pU):

Lower limit (pL) Upper limit (pU)

ν1 F ν1 F
p L = --------------------
- p U = --------------------
-
ν 2 + ν1 F ν2 + ν1 F

where: where:
ν1 = 2x ν1 = 2(x + 1)
ν2 = 2(n − x +1) ν2 = 2(n − x)
x = number of successes x = number of successes
n = number of trials n = number of trials
F = lower α/2 point of F with ν1 F = upper α/2 point of F with ν1
and ν2 degrees of freedom and ν2 degrees of freedom

1-26 MINITAB User’s Guide 2

Confidence Interval and Test of a Proportion Basic Statistics

If you choose to use a normal approximation, MINITAB calculates the confidence

interval as:

where:
p̂ ( 1 – p̂ )
p̂ ± z α ⁄ 2 -------------------- p̂ is the observed probability, p̂ = x / n, where x is the
n observed number of successes in n trials
zα/2 is the value from the z-distribution where α is
1 − confidence level / 100
n is the number of trials

You can specify a confidence level of any number between 1 and 100 in Confidence
level. The confidence level is 95% by default.

Hypothesis test
By default, MINITAB uses an exact method to calculate the test probability. If you
choose to use a normal approximation, MINITAB calculates the test statistic (Z) as:

where:
p̂ – p o
Z = ----------------------------- p̂ is the observed probability, p̂ = x / n, where x is the
po ( 1 – p o ) observed number of successes in n trials
-------------------------
n po is the hypothesized probability
n is the number of trials

The probabilities are obtained from a standard Z-distribution table.

When p0 is not specified in Test proportion, p0 = .5 is used. MINITAB performs a
two-tailed test unless you specify a one-tailed test.

e Example of a confidence interval and test for a proportion

A county district attorney would like to run for the office of state district attorney. She
has decided that she will give up her county office and run for state office if more than
65% of her party constituents support her. You need to test H0: p = .65 versus
H1: p > .65.
As her campaign manager, you collected data on 950 randomly selected party
members and find that 560 party members support the candidate. A test of proportion
and confidence interval was performed to determine whether or not the proportion of
supporters was greater than the required proportion of .65.
1 Choose Stat ➤ Basic Statistics ➤ 1 Proportion.

2 Choose Summarized data.

3 In Number of trials, enter 950. In Number of successes, enter 560.

4 Click Options.

MINITAB User’s Guide 2 1-27

Chapter 1 Confidence Interval and Test of Two Proportions

5 In Test proportion, enter 0.65.

6 From Alternative, choose greater than. Click OK twice.

Session Test and Confidence Interval for One Proportion

window
output Test of p = 0.65 vs p > 0.65
Exact
Sample X N Sample p 95.0 % CI P-Value
1 560 950 0.589474 (0.557428, 0.620963) 1.000

Interpreting the Results

The p-value of 1.0 suggests that the data are consistent with the null hypothesis (H0: p =
.65), that is, the proportion of party members that support the candidate is not at
different from .65. You would conclude that the proportion of party members that
support the candidate (p = .59) is not greater than the required proportion of .65. As her
campaign manager, you would advise her not to run for the office of state district
attorney.

Confidence Interval and Test of Two Proportions

Use the 2 Proportions command to compute a confidence interval and perform a
hypothesis test of the difference between two proportions. For example, suppose you
wanted to know whether the proportion of consumers who return a survey could be
increased by providing an incentive such as a product sample. You might include the
product sample with half of your mailings and see if you have more responses from the
group that received the sample than from those who did not. For a two-tailed test of two
proportions:
H0: p1 = p2 versus H1: p1 ≠ p2 where p1 and p2 are the proportions of success
in populations 1 and 2, respectively.
To test one proportion, use Stat ➤ Basic Statistics ➤ 1 Proportion described on page
1-24.

Data
Data can be in two forms: raw or summarized.

Raw data
Raw data can be entered in two ways: stacked and unstacked.
■ enter both samples in a single column (stacked) with a group column to identify the
population. Columns may be numeric, text, or date/time. Successes and failures are

1-28 MINITAB User’s Guide 2

Confidence Interval and Test of Two Proportions Basic Statistics

determined by numeric or alphabetical order. MINITAB defines the lowest value as

the failure; the highest value as the success. For example:
– for the numeric column entries of “5” and “10,” observations of 5 are considered
failures; observations of 10 are considered successes.
– for the text column entries of “agree” and “disagree,” observations of agree are
considered failures; observations of disagree are considered successes. If the data
entries are “yes” and “no,” observations of no are considered failures;
observations of yes are considered successes.
■ enter each sample (unstacked) in separate numeric or text columns. Both columns
must be the same type—numeric or text. Successes and failures are defined as above
for stacked data.
You can reverse the definition of success and failure in a text column by applying a
value order—see Ordering Text Categories in the Manipulating Data chapter of
MINITAB User’s Guide 1.
The sample sizes do not need to be equal. MINITAB automatically omits missing data
from the calculations.

Summarized data
Enter the number of trials and the number of successes for each sample directly in the
2 Proportions dialog box.

h To calculate a confidence interval and test the difference in proportions

1 Choose Stat ➤ Basic Statistics ➤ 2 Proportions

2 Do one of the following:

■ If your raw data are stacked in a single column:

1 Choose Samples in one column.
2 In Samples, enter the column containing the raw data.
3 In Subscripts, enter the column containing the group or population codes.

MINITAB User’s Guide 2 1-29

Chapter 1 Confidence Interval and Test of Two Proportions

■ If your raw data are unstacked, that is, each sample is in a separate column:
1 Choose Samples in different columns.
2 In First, enter the column containing the first sample.
3 In Second, enter the column containing the other sample.

■ If you have summarized data:

1 Choose Summarized data.
2 In Number of trials, enter a numeric value for each sample.
3 In Number of successes, enter a numeric value for each sample.

4 If you like, use one or more of the options listed below, and click OK.

Options

Options subdialog box

■ specify a confidence level for the confidence interval. The default is 95%.
■ specify a null hypothesis test difference. The default is 0.
■ define the alternative hypothesis by choosing less than (lower-tailed), not equal
(two-tailed), or greater than (upper-tailed). The default is a two-tailed test.
■ use a pooled estimate of p to calculate the test statistic. See Hypothesis test on page
1-31.

Method

Confidence interval
The confidence interval is calculated as

p̂ 1 ( 1 – p̂ 1 ) p̂ 2 ( 1 – p̂ 2 )
p̂ 1 – p̂ 2 ± z α ⁄ 2 ------------------------
- + -------------------------
n1 n2

where:
p̂ 1 and p̂ 2 are the observed probabilities of sample one and sample two
respectively, p̂ = x / n, where x is the observed success in n trials
zα/2 is the value from a Z-distribution where α is 1 − confidence level / 100
You can specify a confidence level of any number between 1 and 100 in Confidence
level. The confidence level is 95% by default.

1-30 MINITAB User’s Guide 2

Confidence Interval and Test of Two Proportions Basic Statistics

Hypothesis test
The calculation of the test statistic, Z, depends on the method used to estimate of p. By
default, MINITAB uses separate estimates of p for each population and calculates Z by:

( p̂ 1 – p̂ 2 ) – d o
Z = -------------------------------------------------------------
-
p̂ 1 ( 1 – p̂ 1 ) p̂ 2 ( 1 – p̂ 2 )
------------------------- + -------------------------
n1 n2

where d0 is the hypothesized difference. When d0 is not specified in Test difference, d0

= 0 is used. MINITAB performs a two-tailed test unless you specify a one-tailed test.
If you choose to use a pooled estimate of p for the test, MINITAB calculates Z by:

p̂ 1 – p̂ 2
Z = ---------------------------------------------------
-
 1
p̂ c ( 1 – p̂ c ) ----- + -----
1
n 
1 n2

where p̂ c is the pooled estimate of p (pooled observed probability).

x1 + x2
p̂ c = ----------------
-
n1 + n2

MINITAB performs a two-tailed test unless you specify a one-tailed test. You can only
use this option when the hypothesized difference is zero (d0 = 0).

e Example of a confidence interval and test of two proportions

As your corporation’s purchasing manager, you need to authorize the purchase of
twenty new photocopy machines. After comparing many brands in terms of price, copy
quality, warranty, and features, you have narrowed the choice to two: Brand X and
Brand Y. You decide that the determining factor will be the reliability of the brands as
defined by the proportion requiring service within one year of purchase.
Because your corporation already uses both of these brands, you were able to obtain
information on the service history of 50 randomly selected machines of each brand.
Records indicate that six Brand X machines and eight Brand Y machines needed
service. Use this information to guide your choice of brand for purchase.
1 Choose Stat ➤ Basic Statistics ➤ 2 Proportions.

2 Choose Summarized data.

3 In First sample, under Trials, enter 50. Under Successes, enter 44.

4 In Second sample, under Trials, enter 50. Under Successes, enter 42. Click OK.

MINITAB User’s Guide 2 1-31

Chapter 1 Correlation

Session Test and Confidence Interval for Two Proportions

window
output Sample X N Sample p
1 44 50 0.880000
2 42 50 0.840000

Estimate for p (1) - p (2): 0.04

95 % CI for p (1) - p (2): (-0.0957903, 0.175790)
Test for p (1) - p (2) = 0 (vs not = 0): Z = 0.58 P-Value = 0.564

* NOTE * The normal approximation may be inaccurate for small samples.

Interpreting the results

Since the p-value of 0.564 is larger than likely α levels, the data are consistent with the
null hypothesis (H0: p1 = p2), that is, the proportion of photocopy machines that
needed service in the first year did not differ depending on brand. As the purchasing
manager, you need to find a different criterion to guide your decision on which brand
to purchase.
You can make the same decision using the 95% confidence interval. Because zero falls
in the confidence interval (−0.096, 0.176), you can conclude that the data are
consistent with the null hypothesis. If you think that the confidence interval is too wide
and does not provide precise information as to the value of p1 − p2, you may want to
collect more data in order to obtain a better estimate of the difference.

Correlation
You can use the Pearson product moment correlation coefficient to measure the degree
of linear relationship between two variables. The correlation coefficient assumes a
value between −1 and +1. If one variable tends to increase as the other decreases, the
correlation coefficient is negative. Conversely, if the two variables tend to increase
together the correlation coefficient is positive. For a two-tailed test of the correlation:
H0: ρ = 0 versus H1: ρ ≠ 0 where ρ is the correlation between a pair of variables.

Data
Data must be in numeric columns of equal length.
MINITAB omits missing data from calculations using a method that is often called
pairwise deletion. MINITAB omits from the calculations for each column pair only those
rows that contain a missing value for that pair.
If you are calculating correlations between multiple columns at the same time,
pairwise deletion may result in different observations being included in the various

1-32 MINITAB User’s Guide 2

Correlation Basic Statistics

correlations. Although this method is the best for each individual correlation, the
correlation matrix as a whole may not be well behaved (for example, it may not be
positive definite).

h To calculate the Pearson product moment correlation

1 Choose Stat ➤ Basic Statistics ➤ Correlation.

2 In Variables, enter the columns containing the measurement data.

3 If you like, use either of the options listed below, then click OK.

Options
■ display the p-value for individual hypothesis tests. This is the default.
■ store the correlation matrix. MINITAB does not display the correlation matrix when
you store the matrix. To display the matrix, choose File ➤ Display Data.

Method
For the two variables x and y,

r = ∑
(x – x)(y – y)
----------------------------------------
( n – 1 )s x s y

where x and sx are the sample mean and standard deviation for the first sample, and y
and sy are the sample mean and standard deviation for the second sample.

e Example of Pearson correlations

We have verbal and math SAT scores and first-year college grade-point averages for 200
students and we wish to investigate the relatedness of these variables. We use
correlation with the default choice for displaying p-values.

MINITAB User’s Guide 2 1-33

Chapter 1 Correlation

1 Open the file [Link].

2 Choose Stat ➤ Basic Statistics ➤ Correlation.

3 In Variables, enter Verbal Math GPA. Click OK.

Session Correlations (Pearson)

window
Verbal Math
output Math 0.275
0.000

GPA 0.322 0.194

0.000 0.006

Interpreting the results

MINITAB displays the correlation for the lower triangle of the correlation matrix when
there are more than two variables. The Pearson correlation between Math and Verbal is
0.275, between GPA and Verbal is 0.322, and between GPA and Math is 0.194.
MINITAB prints the p-values for the individual hypothesis tests of the correlations being
zero below the correlations. Since all the p-values are smaller than 0.01, there is
sufficient evidence at α = 0.01 that the correlations are not zero, in part reflecting the
large sample size of 200.

Spearman’s ρ
You can obtain Spearman’s ρ (rank correlation coefficient) by ranking both columns
and then calculating the correlation for the columns of ranks. See the Manipulating
Data chapter in MINITAB User’s Guide 1.
Before ranking the data, delete any rows that contain missing values.

Partial correlation coefficients

By using a combination of MINITAB commands, you can also compute a partial
correlation coefficient which is a case of correlation. This is the correlation coefficient
between two variables while adjusting for the effects of other variables. Partial
correlation coefficients can be used when you have multiple potential predictors, and
you wish to examine the individual effect of predictors upon the response variable after
taking into account the other predictors.

h To calculate a partial correlation coefficient between two variables

1 Regress the first variable on the other variables and store the residuals. See
Regression on page 2-3.
2 Regress the second variable on the other variables and store the residuals.

3 Calculate the correlation between the two columns of residuals.

1-34 MINITAB User’s Guide 2

Correlation Basic Statistics

e Example of computing a partial correlation coefficient

A survey was conducted in restaurants in 19 Wisconsin counties. Variables measured
include: Sales, the gross sales; Newcap, new capital invested; and Value, estimated
market value of the business. All variables are measured in thousands of dollars.
We want to look at the relationship between sales and new capital invested removing
the influence of market value of the business. First we calculate the regular Pearson
correlation coefficient for comparison. Then we demonstrate calculating the partial
correlation coefficient between sales and new capital.

Step 1: Calculate unadjusted correlation coefficients

1 Open the file [Link].

2 Choose Stat ➤ Basic Statistics ➤ Correlation.

3 In Variables, enter Sales Newcap Value. Click OK.

The remaining steps calculate partial correlation between Sales and Newcap.

Step 2: Regress Sales on Value and store the residuals (Resi1)

1 Choose Stat ➤ Regression ➤ Regression.

2 In Response, enter Sales. In Predictors, enter Value.

3 Click Storage, and check Residuals. Click OK twice.

Step 3: Regress Newcap on Value and store the residuals (Resi2)

1 Choose Stat ➤ Regression ➤ Regression.

2 In Response, enter Newcap. In Predictors, enter Value.

3 Click Storage, and check Residuals. Click OK twice.

Step 4: Calculate correlations of the residual columns

1 Choose Stat ➤ Basic Statistics ➤ Correlation.

2 In Variables, enter Resi1 and Resi2. Click OK.

Session Correlations (Pearson)

window
output Sales Newcap
Newcap 0.615
0.000

Value 0.803 0.734

0.000 0.000
Correlations (Pearson)

Correlation of RESI1 and RESI2 = 0.078, P = 0.261

MINITAB User’s Guide 2 1-35

Chapter 1 Covariance

Interpreting the results

The correlation between the residual columns is 0.078. In other words, after adjusting
for the linear effect of Value, the correlation between Sales and Newcap is 0.078—a
value that is quite different from the uncorrected 0.615 value. In addition, the p-value
of 0.261 indicates that there is no evidence that the correlation between Sales and
Newcap—after accounting for the Value effect—is different from zero.
You can repeat this example to obtain the partial correlation coefficients between other
variables. The partial correlation between Sales and Value is 0.654; the partial
correlation between Newcap and Value is 0.502.

Covariance
You can calculate the covariance for all pairs of columns. Like the Pearson correlation
coefficient, the covariance is a measure of the relationship between two variables.
However, the covariance has not been standardized, as is done with the correlation
coefficient. The correlation coefficient is standardized by dividing by the standard
deviation of both variables.

Data
Data must be in numeric columns of equal length.
MINITAB omits missing data from calculations using a method that is often called
pairwise deletion. MINITAB omits from the calculations for each column pair only those
rows that contain a missing value for that pair.
If you are calculating covariances between multiple columns at the same time,
pairwise deletion may result in different observations being included in the various
covariances. Although this method is the best for each individual covariance, the
covariance matrix as a whole may not be well behaved (for example, it may not be
positive definite).

1-36 MINITAB User’s Guide 2

Normality Test Basic Statistics

h To calculate the covariance

1 Choose Stat ➤ Basic Statistics ➤ Covariance.

2 In Variables, enter the columns containing the measurement data.

3 If you like, use the option listed below, then click OK.

Options
You can store the covariance matrix. MINITAB does not display the covariance matrix
when you store the matrix. To display the matrix, choose File ➤ Display Data.

Method
The covariance between each pair of columns is calculated, using the formula

∑ ( x – x ) ( y – y )-
S xy = ---------------------------------------
n–1
where x is the sample mean for the first sample and y is the sample mean for the
second sample.

Normality Test
Normality test generates a normal probability plot and performs a hypothesis test to
examine whether or not the observations follow a normal distribution. For the
normality test,
H0: data follow a normal distribution vs. H1: data do not follow a normal distribution

MINITAB User’s Guide 2 1-37

Chapter 1 Normality Test

Data
You need one numeric column. MINITAB automatically omits missing data from the
calculations.

h To perform a normality test

1 Choose Stat ➤ Basic Statistics ➤ Normality Test.

2 In Variable, enter the column containing the measurement data.

3 If you like, use one or more of the options listed below, and click OK.

Options
■ mark reference probabilities and corresponding data values on the plot—see
Method on page 1-39
■ perform an Anderson-Darling, Ryan-Joiner, or Kolmogorov-Smirnov test for
normality—see Choosing a normality test below
■ replace the default graph title with your own title

Choosing a normality test

You have a choice of hypothesis tests for testing normality:
■ Anderson-Darling test (the default), which is an ECDF (empirical cumulative
distribution function) based test
■ Ryan-Joiner test [3], [5] (similar to the Shapiro-Wilk test [6], [7]) which is a
correlation based test
■ Kolmogorov-Smirnov test, an ECDF based test
The Anderson-Darling and Ryan-Joiner tests have similar power for detecting
non-normality. The Kolmogorov-Smirnov test has lesser power. See [2] and [5] for
discussions of these tests for normality.
The common null hypothesis for these three tests is H0: data follow a normal
distribution. If the p-value of the test is less than your α level, reject H0.

1-38 MINITAB User’s Guide 2

Normality Test Basic Statistics

Method
The input data are plotted as the x-values. MINITAB calculates the probability of
occurrence, assuming a normal distribution, and plots the calculated probabilities as
y-values. The grid on the graph resembles the grids found on normal probability paper,
with a log scale for the probabilities. A least-squares line is fit to the plotted points and
drawn on the plot for reference. The line forms an estimate of the cumulative
distribution function for the population from which data are drawn. MINITAB also
displays the sample mean, standard deviation, and sample size on the plot.
When you enter the optional reference probabilities, they are marked with horizontal
references lines. At the point where the reference line intersects the least-squares fit, a
vertical reference line is drawn and labeled with the corresponding data value. To
include reference probabilities on the plot:
In Reference probabilities, enter a column containing the reference probabilities,
which must be values between 0 and 1.

e Example of an Anderson-Darling normality test

In an operating engine, parts of the crankshaft move up and down. AtoBDist is the
distance (in mm) from the actual (A) position of a point on the crankshaft to a baseline
(B) position. To ensure production quality, a manager took five measurements each
working day in a car assembly plant, from September 28 through October 15, and then
ten per day from the 18th through the 25th.
You wish to see if these data follow a normal distribution, so you use Normality test.
1 Open the file [Link].

2 Choose Stat ➤ Basic Statistics ➤ Normality Test.

3 In Variable, enter AtoBDist. Click OK.

Graph
window
output

MINITAB User’s Guide 2 1-39

Chapter 1 References

Interpreting the results

The graphical output is a plot of normal probabilities versus the data. The data depart
from the fitted line most evidently in the extremes, or distribution tails. The
Anderson-Darling test’s p-value indicates that, at α levels greater than 0.022, there is
evidence that the data do not follow a normal distribution. There is a slight tendency
for these data to be lighter in the tails than a normal distribution because the smallest
points are below the line and the largest point is just above the line. A distribution with
heavy tails would show the opposite pattern at the extremes.

References
[1] S.F. Arnold (1990). Mathematical Statistics, Prentice-Hall, pp.383-384.

[2] R.B. D’Augostino and M.A. Stevens, Eds. (1986). Goodness-of-Fit Techniques,
Marcel Dekker.

[3] J.J. Filliben (1975). “The Probability Plot Correlation Coefficient Test for
Normality,” Technometrics, Vol 17, p.111.

[4] N.L. Johnson and S. Kotz (1969). Discrete Distributions, John Wiley & Sons,
pp.58-61.

[5] T.A. Ryan, Jr. and B.L. Joiner (1976). “Normal Probability Plots and Tests for
Normality,” Technical Report, Statistics Department, The Pennsylvania State
University. (Available from Minitab Inc.)

[6] S.S. Shapiro and R.S. Francia (1972). “An Approximate Analysis of Variance Test
for Normality,” Journal of the American Statistical Association, Vol 67, p.215.

[7] S.S. Shapiro and M.B. Wilk. “An Analysis of Variance Test for Normality
(Complete Samples),” Biometrika, Vol 52, p. 591.

1-40 MINITAB User’s Guide 2