Module 9: Nonparametric Tests

The Applied Research Center

Module 9 Overview
}  Nonparametric Tests
}  Parametric vs. Nonparametric Tests
}  Restrictions of Nonparametric Tests
}  One-Sample Chi-Square Test
}  Chi-Square Test of Independence
}  Other Nonparametric Tests
What Are Nonparametric Tests?
}  Nonparametric tests require few, if any assumptions
about the shapes of the underlying population
distributions
}  For this reason, they are often used in place of parametric
tests if or when one feels that the assumptions of the
parametric test have been too grossly violated (e.g., if the
distributions are too severely skewed).
Parametric or Nonparametric tests?
}  If all assumptions are met, use Parametric techniques
}  Use Nonparametric techniques
}  When the dependent variable is either nominal or ordinal
}  If the distribution of the dependent variable is skewed
}  When the assumptions are not met, specifically:
}  Normality
}  Homogeneity of variance
Restrictions
}  Nonparametric tests do have at least two major
disadvantages in comparison to parametric tests:
}  First, nonparametric tests are less powerful. Why?
Because parametric tests use more of the information
available in a set of numbers.
}  Parametric tests make use of information consistent with
interval or ratio scale (or continuous) measurement,
whereas nonparametric tests typically make use of
nominal or ordinal (or categorical) information only.

5

Restrictions (cont d)
}  Second, parametric tests are much more flexible, and
allow you to test a greater range of hypotheses. For
example, ANOVA designs allow you to test for
interactions between variables in a way that is not
possible with nonparametric alternatives.
}  There are nonparametric techniques to test for certain
kinds of interactions under certain circumstances, but
these are much more limited than the corresponding
parametric techniques.

6

Chi-Square Test
}  Used to test variables that have nominal data
}  Examples of nominal data
}  Gender
}  Political affiliation
}  Course delivery method
}  Three main types:
}  Goodness-of-fit (or one-sample) test
}  Test of independence (or association)
}  Independent-samples test
Chi-Square Goodness-of-Fit Test
¨  Evaluates whether the proportions of individuals who fall
into categories are equal to hypothesized values
¨  The variable can have two or more categories
¨  The categories can have quantitative (one category
reflects a higher value than another; e.g., Likert scale
responses of Agree and Disagree) or qualitative grouping
(e.g., course delivery method)
¨  Note: The chi-square test does not recognize any
quantitative distinction among categories; it simply
assesses whether the proportions associated with the
categories are significantly different from the
hypothesized values
Chi-Square Goodness-of-Fit Test
(cont d)
}  When conducting a chi-square test in SPSS, you must first
specify the values for the hypothesized proportions
}  The proportions can be
}  Equal (proportion = 1/ number of categories)
}  Not equal (determined based on previous research)
Chi-Square Goodness-of-Fit Test
(cont d)
¨  Next, you must formulate a null hypothesis (the expected
frequencies = hypothesized proportions)
¨  Then, specify the expected frequencies associated with
the categories.
¤  The expected frequency =
hypothesized proportion x total sample size.
¤  If hypothesized proportions are equal, the expected
frequencies will be equal
¤  If hypothesized proportions are unequal, the expected
frequencies will be unequal
Contingency Tables
}  Helps arrange the data in a study for the purpose of
showing how each group split itself up on the variable
}  Make it easier to understand the null hypothesis and why
the data led to the rejection or not of the null.
Goodness-of-Fit Example
¨  A professor was interested in whether students prefer to
take classes face to face or online. A poll was taken with
100 students to determine their preference for learning.
¨  Step 1: State the null and alternate hypotheses
¤  H0 : The number of students who prefer to take classes face to
face is equal to the number of students who prefer to take
classes online
¤  Ha : The number of students who prefer to take classes face to
face is not equal to the number of students who prefer to
take classes online
Goodness-of-Fit Example (cont d)
}  Step 2: Create the contingency table
Observed Hypothesized Expected
Frequencies Proportions Frequencies

Face to Face 39 .50 50

Online 61 .50 50

}  Step 3: Input the data into SPSS

}  Create a variable labeled DeliveryMethod
}  Assign values: 1 = Face to Face, 2 = Online
}  Enter the data (in one column): 39, 1 s and 61, 2 s
Goodness-of-Fit Example (cont d)
}  Step 4: Run the Analysis
}  Analyze à Nonparametric à Legacy Dialogs à
Chi-square
}  Move the variable (DeliveryMethod) to the Test variable list
box
}  Under Expected Values, select All categories equal
}  Click Ok
Goodness-of-Fit Example (cont d)
Goodness-of-Fit Example (cont d)
}  Step 4: Make a decision regarding the null
}  X2 = 4.84
}  df = 1 (Number of categories – 1)
}  p = .028
}  What is our decision regarding the null?
Goodness-of-Fit Example (cont d)
}  Using the level of significance = .05, do we reject or fail to
reject the null?
}  If p < .05, we reject the null
}  if p > .05, we fail to reject the null
}  According to SPSS, p = .028
}  .028 < .05, therefore, we reject the null!
Goodness-of-Fit Example (cont d)
}  Step 5: Write up your results.
}  The null hypothesis stated that the number of students who
prefer to take classes face to face is equal to the number of
students who prefer to take classes online. A chi-square
goodness-of-fit test revealed that significantly more students
prefer to take courses online (61%) compared to face to face
(39%, X2 = 4.84, df = 1, p = .028). Consequently, the null
hypothesis was rejected.
Chi-Square Test of Independence
}  A test to see whether a relationship (or association)
exists between two nominal variables
}  Similar to a bivariate correlation
Test of Independence Example
}  A professor is interested in whether there is a
relationship between gender and preference for learning
statistics face to face or online. A poll was taken with 100
students (50 males and 50 females) to determine their
preference for learning.
}  Step 1: State the null and alternate hypotheses
}  H0 : There is no relationship between gender and preference
for learning statistics face to face or online
}  Ha : There is a relationship between gender and preference for
learning statistics face to face or online
Test of Independence Example (cont d)
}  Step 2: Create the contingency table
Face to face Online
Males 20 30
Females 30 20

}  Step 3: Input the data into SPSS

}  Create 2 variables: Gender and DeliveryMethod
}  Assign values
}  Enter the data:
}  For Gender: 50, 1 s and 50, 2 s
}  For DeliveryMethod: 20 males and 30 females (face to face), 30 males
and 20 females (online)
Test of Independence Example (cont d)
}  Step 4: Run the Analysis
}  Analyze à Descriptives à Crosstabs
}  Move Gender to the Row box and DeliveryMethod to the
Column box
}  Select Statistics then select Chi square and Phi and
Cramer s V boxes, Click Continue
}  Select Cells then under Counts select Observed and
Expected, Click Continue
}  Select Display clustered bar charts at the bottom left.
}  Click Ok
Test of Independence Example (cont d)
Test of Independence Example (cont d)
Test of Independence Example (cont d)
Test of Independence Example (cont d)
}  Step 4: Make a decision regarding the null
}  X2 = 4.00
}  df = 1
}  (Number of categories in 1st variable – 1)(Number of categories in 2nd
variable – 1)
}  p = .046
}  What is our decision regarding the null?
Test of Independence Example (cont d)
}  Using the level of significance = .05, do we reject or fail to
reject the null?
}  If p < .05, we reject the null
}  if p > .05, we fail to reject the null
}  According to SPSS, p = .046
}  .046 < .05, therefore, we reject the null!
Test of Independence Example (cont d)
}  Step 5: Write up your results.
}  The null hypothesis stated that there is no relationship
between gender and preference for learning statistics face to
face or online. A chi-square test of independence revealed a
significant relationship between gender and preference for
learning statistics, X2 (1, N = 100) = 4.00; p = .046.
Consequently, the null hypothesis was rejected.
Nonparametric Tests for Nominal Data
}  The sign test
}  The binomial test
}  Fischer s Exact Test
}  The chi-square test
}  McNemar s test
}  Cochran s test
Nonparametric Tests for Ordinal Data
}  The Mann-Whitney U test
}  Analogous to independent samples t test
}  Wilcoxon signed-ranks test
}  Analogous to paired samples t test
}  The Kruskal-Walis one-way analysis of ranks
}  Analogous to a one-way ANOVA
}  The Friedman s ANOVA
}  Analogous to RM ANOVA
}  Spearman s rho
}  Analogous to Pearson correlation
Module 9 Summary
}  Nonparametric Tests
}  Parametric vs. Nonparametric Tests
}  Restrictions of Nonparametric Tests
}  One-Sample Chi-Square Test
}  Chi-Square Test of Independence
}  Other Nonparametric Tests
Review Activity
}  Please complete the review activities at the end of the
module.
}  You can complete the review activities and module as
many times as you like.
}  Upon completion of the review activities you will have an
opportunity to print your Certificate of Completion,
along with instructions for submitting your certificate to
the ARC.
Statistics PD Course Modules
}  Module 1: Introduction to Statistics
}  Module 2: Introduction to SPSS
}  Module 3: Descriptive Statistics
}  Module 4: Inferential Statistics
}  Module 5: Correlation
}  Module 6: t-Tests
}  Module 7: ANOVAs
}  Module 8: Linear Regression
}  Module 9: Nonparametric Procedures