STAT 20053

STATISTICAL M6: NON-PARAMETRIC STATISTICS

ANALYSIS with
SOFTWARE
APPLICATION

RUMEL ANGELO T. ALFARO, MSc. (Cand.), CKSSBB

P U P 1
 COURSE OUTCOME & TOPICS

Where we’re going… Topics

STAT 20053
• Identify the difference between • Parametric vs Non-Parametric Tests (6.1)
STATISTICAL parametric and non-parametric • Mann-Whitney U Test (6.2)
ANALYSIS with statistical methods.
SOFTWARE • Test hypotheses, using the Mann- • Wilcoxon Signed-Rank Test (6.3)
APPLICATION Whitney U test • Kruskal-Wallis Test (6.4)
• Test hypotheses, using the signed-
rank test. • Chi-square Test (6.5)
• Test hypotheses, using the
Kruskal-Wallis test.
• Test hypotheses on contingency
anlaysis, using the Chi-square test.
P U P
2
 PARAMETRIC VS
NON-PARAMETRIC TESTS

Parametric Non-Parametric
STAT 20053 Evaluate hypotheses for a particular Evaluate hypotheses for the entire
parameter, usually the population mean. population distribution.
STATISTICAL
Quantitative data Quantitative, ranked, qualitative data
ANALYSIS with
SOFTWARE Require assumptions about the Require no assumptions (“distribution-
APPLICATION distributional characteristics of the free”) so used with non-normal
population distribution (e.g., normal distributions and when variances of the
distribution, homogeneity of variances) groups are not equal.
More powerful than non-parametric Generally use to compute.
tests when assumptions are met.

P U P
3
 PARAMETRIC VS
NON-PARAMETRIC TESTS

STAT 20053
Parametric Non-Parametric
STATISTICAL Two Independent Samples Independent t-Test Mann-Whitney U
ANALYSIS with
Dependent/Paired Samples Paired t-Test Wilcoxon Signed-Rank
SOFTWARE
APPLICATION More than Two Samples One-way ANOVA Kruskal-Wallis
Nominal Samples N/A Chi-Square

P U P
4
 MANN-WHITNEY U TEST

The Mann-Whitney U Test (sometimes called Wilcoxon rank-sum test) is a

nonparametric test that uses ranks of sample data from two independent populations.
STAT 20053 It is used to test the null hypothesis that the two independent samples come from
populations with equal medians. The alternative hypothesis is the claim that the two
STATISTICAL populations have different medians.
ANALYSIS with 𝐻0 : The two samples come from populations with equal medians.
SOFTWARE 𝐻1 : The two samples come from populations with different medians.
APPLICATION
Requirements
1. There are two independent samples of randomly selected data measured at a
continuous level (i.e., they are interval or ratio variables) or ordinal level.
2. The test works best for two samples with more than 10 values.
3. There is no requirement that the two populations have a normal distribution or
any other particular distribution.

P U P
5
 MANN-WHITNEY U TEST

Test Statistic
STAT 20053 𝑈=
𝑅 − 𝜇𝑅
𝜎𝑅
STATISTICAL where
𝑛1 𝑛1 + 𝑛2 + 1
ANALYSIS with 𝜇𝑅 =
2
SOFTWARE
𝑛1 𝑛2 𝑛1 + 𝑛2 + 1
APPLICATION 𝜎𝑅 =
12

• 𝑅 = same as 𝑅1 (sum of ranks for Sample 1)

• 𝜇𝑅 = mean of the sample R values that is expected when the two populations have equal medians
• 𝜎𝑅 =standard deviation of the sample R values that is expected with two populations
• having equal medians

P U P
6
 MANN-WHITNEY U TEST

STAT 20053 Steps in Minitab:

1. Test for normality of the distribution, if needed.
STATISTICAL 2. Open the Mann-Whitney Test dialog box.
ANALYSIS with • Mac: Stat > 2-Sample Inference > Mann-Whitney
SOFTWARE
• PC: Stat > Non-Parametrics > Mann-Whitney
APPLICATION
3. From the drop-down list, select Each sample is in its own column.
4. In First Sample, enter Group 1 data.
5. In Second Sample, enter Group 2 data.
6. Click OK.

P U P
7
 MANN-WHITNEY U TEST

STAT 20053 Illustrative Example

1. Two independent samples of army and marine recruits are selected, and the
STATISTICAL time in minutes it takes each recruit to complete an obstacle course is
ANALYSIS with recorded, as shown in the table. At 𝛼 = 0.05, is there a difference in the
SOFTWARE times it takes the recruits to complete the course?
APPLICATION
Army 15 18 16 55 13 2 24 17 19 21 26 28 46
Marines 14 9 16 19 88 12 11 8 74 18 25

P U P
8
 MANN-WHITNEY U TEST

STAT 20053 Illustrative Example

2. When the author visited Dublin in Ireland, he recorded the ages of randomly
STATISTICAL selected passenger cars and randomly selected taxis. The ages (in years) are
ANALYSIS with listed below. Use a 0.05 significance level to test the claim that there is a
SOFTWARE difference between the median age of a Dublin car and the median age of a
APPLICATION Dublin taxi. We might expect that taxis would be newer, but what do the
results suggest?
Data:Ages

P U P
9
 WILCOXON SIGNED-RANK TEST

When the samples are dependent, as they would be in a before-and-after test using
the same subjects, the Wilcoxon signed-rank test can be used in place of the t test for dependent
STAT 20053 samples. Again, this test does not require the condition of normality. It uses the following
hypotheses:
𝐻0 : The matched pairs have differences that come from a population with a median equal to zero.
STATISTICAL
ANALYSIS with 𝐻1 : The matched pairs have differences that come from a population with a nonzero median.
SOFTWARE
APPLICATION Test Statistic
𝑛(𝑛 + 1)
𝑤𝑠 −
𝑧= 4
𝑛(𝑛 + 1)(2𝑛 + 1)
24
where
• 𝑛 = number of pairs where difference is not 0
• 𝑤𝑠 = smaller sum in absolute value of signed ranks
P U P
10
 WILCOXON SIGNED-RANK TEST

Requirements
1. Your dependent variable should be measured at a continuous level (i.e., they are interval or ratio
STAT 20053 variables) or ordinal level.
2. Your independent variable should consist of two categorical, "related groups" or "matched pairs".
STATISTICAL 3. There is no requirement that the two populations have a normal distribution or any other
ANALYSIS with particular distribution.
SOFTWARE
APPLICATION Steps in Minitab:
1. Test for normality of the distribution, if needed.
2. The difference scores that you need when running a Wilcoxon signed-rank test in Minitab are not
automatically calculated. Therefore, you need to run the Calc > Calculator... procedure.
3. Click Stat > Nonparametrics > 1-Sample Wilcoxon
4. Input the needed variables, etc.
5. Click OK
P U P
11
 WILCOXON SIGNED-RANK TEST

STAT 20053 Illustrative Example

1. In a large department store, the owner wishes to see whether the number of
STATISTICAL shoplifting incidents per day will change if the number of uniformed security
ANALYSIS with officers is doubled. A sample of 7 days before security is increased and 7 days
SOFTWARE after the increase shows the number of shoplifting incidents. Is there enough
APPLICATION evidence to support the claim, at 𝛼 = 0.10, that there is a difference in the
number of shoplifting incidents before and after the increase in security?
Data: Security

P U P
12
 WILCOXON SIGNED-RANK TEST

STAT 20053 Illustrative Example

2. A statistics professor wants to investigate the relationship between a
STATISTICAL student’s midterm examination score and the score on the final. Eight
ANALYSIS with students were selected, and their scores on the two examinations are noted
SOFTWARE below. At the 0.10 level of significance, assuming that the difference between
APPLICATION scores is not normal, is there sufficient evidence to conclude that the
students performed better in Final than Midterm exam?
Data: Scores

P U P
13
 KRUSKAL-WALLIS TEST

The analysis of variance uses the F test to compare the means of three or more
populations. The assumptions for the ANOVA test are that the populations are normally
STAT 20053 distributed and that the population variances are equal. When these assumptions cannot be met,
the nonparametric Kruskal-Wallis test, sometimes called the H test, can be used to compare
three or more means.
STATISTICAL
𝐻0 : The samples come from populations with equal medians.
ANALYSIS with
SOFTWARE 𝐻1 : The samples come from populations with medians that are not all equal.
APPLICATION Test Statistic
12 𝑅12 𝑅22 𝑅𝑘2
𝐻= + + ⋯+ − 3(𝑁 + 1)
𝑁 𝑁+1 𝑛1 𝑛2 𝑛𝑘
where
• 𝑁 = total number of observations in all samples combined
• 𝑘 = number of samples
• 𝑅 = sum of ranks for Sample k calculated with the procedure that follows
• 𝑛 = number of observations in Sample k
P U P
14
 KRUSKAL-WALLIS TEST

Requirements
STAT 20053 1. We have at least three independent samples, all of which are randomly selected.
2. This test works best if each sample has at least five observations.
STATISTICAL 3. There is no requirement that the populations have a normal distribution or any
other particular distribution.
ANALYSIS with
SOFTWARE
APPLICATION Steps in Minitab:
1. Test for normality of the distribution, if needed.
2. Click Stat > Nonparametrics > Kruskal-Wallis
3. Select Responses are in one column for all factor levels
4. Input the needed variables, etc.
5. Click OK

P U P
15
 KRUSKAL-WALLIS TEST

STAT 20053 Illustrative Example

1. A researcher tests three different brands of breakfast drinks to see how many
STATISTICAL milliequivalents of potassium per quart each contains. These data are obtained. At
ANALYSIS with 𝛼 =0.10, is there enough evidence to reject the hypothesis that all brands contain
SOFTWARE the same amount of potassium?
APPLICATION Data: Drinks
2. A recent study recorded the number of job offers received by newly graduated
chemical engineers at three colleges. The data are shown here. At 𝛼 =0.05, is
there a difference in the average number of job offers received by the graduates at
the three colleges? Assume that the data is not normally distributed.
Data: Job

P U P
16
 CHI-SQUARE TEST

When data can be tabulated in table form in terms of frequencies, several types of
hypotheses can be tested by using the chi-square test.
STAT 20053
The Chi-Square tests are tests based on hypothesis-testing problems arise for
frequency, or count data. These are tests that are concerned with the testing of statistical
STATISTICAL hypotheses about multiple population proportions.
ANALYSIS with A multinomial experiment is an experiment that meets the following conditions:
SOFTWARE 1. The number of trials is fixed and trials are independent.
APPLICATION 2. Each trial must have all outcomes classified into exactly one of several different
categories.
3. For a 2 by 2 table, all expected frequencies > 5.
Hypotheses:
𝐻0 : The row and the column criteria are independent of each other.
𝐻1 : The row and the column criteria are not independent of each other.

P U P
17
 CHI-SQUARE TEST

Test Statistic

STAT 20053 The formula for the test value for the independence test is as follows:
𝑘
𝑜𝑖 − 𝑒𝑖 2

STATISTICAL 𝜒2 = ෍
𝑒𝑖
ANALYSIS with 𝑖=1

SOFTWARE The symbols 𝑜𝑖 and 𝑒𝑖 represent the observed and expected frequencies, for
APPLICATION the ith cell.
The chi-square independence test can be used to test the
independence of two variables for data expressed in contingency table.
Group Category A Category B
A 𝑜1 𝑜2
B 𝑜3 𝑜4
P U P
18
 CHI-SQUARE TEST

STAT 20053 Steps in Minitab:

1. Choose Stat > Tables > Chi-Square Test
STATISTICAL for Association
ANALYSIS with
2. In Columns and Rows, enter the
SOFTWARE
corresponding variables necessary
APPLICATION
3. Click Statistics > Uncheck Expected Cell
Counts
4. Click OK

P U P
19
 CHI-SQUARE TEST

STAT 20053 Illustrative Example

1. A company has to choose among three health insurance plans. Management
STATISTICAL wishes to know whether the preference for plans is independent of job
classification and wants to use 𝛼 = 0.05. The opinions of a random sample of 20
ANALYSIS with employees are shown. Is the preference in the insurance plan independent to job
SOFTWARE classification?
APPLICATION Data: Insurance
2. Suppose a new postoperative procedure is administered to a number of patients
in a large hospital. The researcher can ask the question, at 𝛼 = 0.05, do the
doctors feel differently about this procedure from the nurses, or do they feel
basically the same way?
Data: Procedure

P U P
20
STAT 20053

STATISTICAL QUESTIONS?
ANALYSIS with
SOFTWARE
APPLICATION
You may reach at the ff channels during Worksheet
Consultation Hours:

• Google Classroom
• Facebook Group
• Discord
P U P
21