Advanced Statistics for Business

Handout for Module 01

Parametric Methods:

Parametric comes from the word parameter. Parameter means numbers that summaries the

characteristics of a population. Parameters are Population mean (μ), Population proportion (p) and
Population variance (σ2). The random observations follow specific probability distributions, that can be
defined by the parameters. The assumption for parametric method is that the data is normally
distributed.
Non - Parametric Methods:

They do not have any assumptions about the underlying distribution of the data. Non - Parametric
Methods are also called as Distribution free methods. Non-parametric methods are valuable because
they make fewer assumptions and don't require data to follow a specific distribution, making them more
flexible and robust in various situations.
Mann-Whitney U Test:

It is also called as Wilcoxon Rank-Sum Test. Non-parametric tests, like the Wilcoxon test, use the
relative ranks of observations to compare groups and determine if one tends to have higher or lower
scores. While more flexible and robust, they require more data to achieve the same precision as
parametric tests.
Concept of Independence

Independence means the occurrence of one event does not affect the occurrence of another. If two
events are independent, knowing one event has occurred gives no information about the other event.
The Multiplication Rule in Probability states that two events, A and B, are independent if
P(A∩B)=P(A)⋅P(B). This means the probability of both events occurring together is the product of their
individual probabilities, with no influence from one event on the other.
Steps to test independence:

1. Collect data on the joint purchases of digital and physical goods by the customers.
2. Calculate the joint probability distribution of X&Y for all combination of values X&Y
P(X=x, Y=y)
3. Compare this joint distribution to the product of the marginal distributions
P(X=x) P(Y=y)
4. If the joint distribution is approximately equal to the product of the marginal distributions, then the
company may conclude that the random variables are independent.
Chi – Squared test:

1. To test whether the two variables are independent of each other.

2. To identify the degrees of freedom for the chi squared variable under the null hypothesis. (r-1)(c-1). r
and c = the number of cases for the two variables.
3. Excel formula used: [Link]( ). The formula for this chi-square statistic is:

Oij = observed frequency

Eij = expected frequency
Excel formulas used in Testing independence:
1. UNIQUE
2. COUNTIFS
3 Key Insights the Chi-Square Test Provides for Numerical Data:

1. It is effective even with non-normally distributed, multi-modal data, or data with outliers

2. It can detect non-linear relationships

3. It is less sensitive to data quality issues or errors in data entry

Goodness of fit test:

It is a statistical method used to determine how well a set of observed data matches a particular theoretical or expected
distribution. Checks the pattern of data expected.

Example: We may expect a Normal distribution or a Poisson distribution, or another type of pattern
Therefore, we need to assess whether our sample observations follow that specific distribution

Objective:

• To check the deviations between the observed frequencies and the expected frequencies are due to random chance or
there is a significant difference between them
• Determine whether the deviations are significant enough to reject the hypothesis (i.e.., the data follows the expected
distribution)
Uses of Poisson distribution:
• When modelling the number of events or occurrences happen within a fixed interval of time or space (especially
during independent events and at a constant average rate)

Scenarios Poisson distribution used in:

• Counting the number of customer arrivals at a store
• The frequency of equipment breakdowns in manufacturing
• The number of projects completed in a company
• The number of defects in production batches
• When a variable that takes integer values such as, 0,1, 2, 3, etc.,

Excel function: [Link]

Goodness of fit test to test the Assumption of Normal Distribution

Two parameters to characterize Normal distribution:

1. The population mean (which determines the center of the distribution)
2. The population standard deviation (which describes the spread of the data)

Objective: To compare the observed data with the expected values from a normal distribution with specified
parameters

Excel function used : [Link]

Wilcoxon signed-rank test
Non – Parametric approach used in one/two sample problems

Advantage:
• It doesn't require any assumptions about the distribution of data.
• It is helpful when the data is not normally distributed, or the sample size is too small to confirm normality.
• It tests if the data in a sample deviate from a specified median, either positively or negatively.
• The test focuses on the ranks of the differences between each data point observed and the hypothesized median.
• It helps to determine if the sample median differs significantly from the expected median, without being affected
by outliers or extreme values.
Wilcoxon signed-rank test

•The test evaluates the magnitude and direction of differences between observed values and the hypothesized median
by using ranks rather than raw data points.

•This approach minimizes the impact of outliers and provides more reliable results for non-normally distributed data.
Wilcoxon signed-rank test - Steps:
[Link] each data point, calculate the difference between the observed value and the hypothesized median.
[Link] differences are ranked by their absolute values, disregarding whether they are positive or negative, and tied
ranks are given the average rank.
[Link] assigning ranks, you add signs (+ or -) based on the direction of the original differences and then calculate
the test statistic as the sum of these signed ranks. If the differences between the observed values and the
hypothesized median are predominantly positive or negative, it indicates a significant difference between the
sample median and the hypothesized median. If the signed ranks balance out, it suggests the sample median is
close to the hypothesized median. Test statistic: Positive signed ranks and denoted as W +; Negative signed ranks
and denote as W-. n=Sample size
Wilcoxon signed-rank test - Steps:
4. In a large sample framework, the test statistic W+ can be used to compute a z-score using a specific formula.
To determine if the test statistic significantly differs from the null hypothesis, you can compare it to a critical value or
calculate the p-value.

•After calculating the z-score, compare it with the critical value from the normal distribution.
•At a 5% significance level, the critical value is 1.96.
•If the absolute value of the z-score exceeds 1.96, you should reject the null hypothesis.
•Decision rule: Reject H0 if |z-score| > 1.96
•Alternatively, you can find the p-value by computing, 2*P (N(0,1) > |z-score|)
•This can be obtained in Excel using [Link]
•Decision rule: Reject H0 if |z-score| > 1.96
Paired Wilcoxon signed-rank test:

• This nonparametric test is used to compare two related samples or paired observations, particularly when the normality
assumption for a t-test is not met.
• It builds on the one-sample rank-sum test and focuses on the median difference between the pairs.
• The test assesses whether this median difference is significantly different from zero.