Psy7620 U10a1 Chi Square

Running head: u10a1 CHI SQUARE
Psy7620: u10a1 Chi Square

Anthony Rhodes
Capella University
2911 Hamilton Blvd. 444

Sioux City, Iowa 51104
Telephone: 712-301-9258
Email: arhodes5@capellauniversity.edu
Instructor: Dr. Art Bangert, PhD.
Chi-Square Test Analysis and Interpretation

Section I: Data File Descriptions
As in previous assignments (Rhodes, 2011), this author used Howell's dataset entitled Add.dat.
This dataset has a sample population of 88 9th grade students derived from a population of 386 children
who during childhood had or had not exhibited symptoms of attention deficit disorder (ADD) (Howell and
Huessy, 1985). Teachers of second-grade school children in various schools in northwestern Vermont were
asked to complete a questionnaire for each of their students dealing with behaviors commonly associated
with ADD. Questionnaires on these same children were repeated in fourth and fifth grades.
The higher the score, the more ADD-like behaviors are exhibited by the child. Information on
these children was obtained from school records at the end of their ninth and twelfth grade. For the dataset
only, the three scores were averaged to obtain a variable identified as ADDSC.
This statistical report will provide a Chi-Square test to examine two independent variables to
determine whether the variables are dependent or independent of each other. According to Howell
(2008), the Chi-Square test is a statistical test used for categorical data or discrete, nominal data which
represent counts or the number of observations in each category under investigation. These counts are
commonly arranged in a tabular format known as a contingency table.
For this discussion, the contingency table is two dimensional so each observation is classified
on the basis of two variables simultaneously. Therefore, the Chi-Square test is designed to test the null
hypothesis that there is no association between the rows and columns of the contingency table or that the
observed counts do not deviate from the expected counts. The independent variables in this Chi-Square

analysis are social problems (SOCPROB) and dropout rate (DROPOUT) for students in the 9th grade
population defined above.
The PASW Statistics GradPack 18 software was used in the following chi-square analysis to
create the Case Processing Summary, Cross Tabulation table, Pearson Chi-Square tests, Phi and Cramers V
Symmetric Measures table, and the Bar Chart. According to the Case Processing Summary in Table 1 for
the independent variables SOCPROB and DROPOUT for the sample population of 9th grade students (N =
88), there were no missing cases.
Table 1
Case Processing Summary for SOCPROB and DROPOUT
Case Processing Summary
Cases
Valid
SOCPROB * DROPOUT
Missing
Percent
88
100.0%
Total
Percent
.0%
Percent
88
100.0%

Note: The above is the Case Processing Summary for SOCPROB and DROPOUT. Adapted from
ADDSC scores [Data file], by Howell, David C., 2009, Statistical Home Page Web site, available
from http://www.uvm.edu/~dhowell/StatPages/
Section II: Assumptions, Data Screening and Verification of Assumptions
A. Since the Chi-Square test is a non-parametric test which does not assume a normal
distribution of data, the assumption for normality is not a significant consideration for this study.
However, the chi-square test assumes that the expected frequencies are greater than 5 to be considered
ideal, with no expected frequencies less than 1. This assumption was verified using the Cross Tabulation
contingency table in Table 2 below. As indicated in the Expected Counts (EC) under both variables:
SOCPROB and DROPOUT, the ECs: 69.1, 8.9 and 8.9 are greater than 5 and therefore considered
ideal. However, the EC in the Yes cell under variables SOCPROB and DROPOUT is 1.1 and less
than 5 and cannot be considered ideal. Therefore the assumption of expected frequencies was violated.
Table 2
Cross Tabulation for variables SOCPROB and DROPOUT
SOCPROB * DROPOUT Cross Tabulation
DROPOUT
No
SOCPROB
No
Count
78
69.1
8.9
78.0
% within SOCPROB
93.6%
6.4%
100.0%
% within DROPOUT
93.6%
50.0%
88.6%
% of Total
83.0%
5.7%
88.6%
10
8.9
1.1
10.0
% within SOCPROB
50.0%
50.0%
100.0%
% within DROPOUT
6.4%
50.0%
11.4%
% of Total
5.7%
5.7%
11.4%
78
10
88
78.0
10.0
88.0
Count
Expected Count
Total
Total
73
Expected Count
Yes
Yes
Count
Expected Count
Note: The above is the Cross Tabulation table for SOCPROB and DROPOUT. Adapted from
B. It is also assumed in this Chi-Square analysis that the data is categorical with a nominal
scale of measurement. This was determined by researcher analysis of the independent variables from
the dataset within the Cross Tabulation contingency table above (Table 2).
C. According to this researchers analysis of the independence of observation, it was
determined according to Table 2 above that the Total Count for two variables was N = 88 and the % of
total equals 100% indicating the total sample size (N) was used and all observations were only counted
once in the contingency table. Therefore the assumption of independence for this Chi-Square analysis
was met.
Section III: Inferential Procedure, Hypothesis and Alpha Level
The research question under investigation is: Is there a relationship between social problems
and dropout rates among the 9th grade student population? The null hypothesis (H0) is: There is no
relationship between social problems and dropout rates among the 9th grade student population. The
alternative hypothesis (Ha) is: There is a relationship between social problems and dropout rates among
the 9th grade student population. The significance level is = .05. The Chi-Square test is two tailed,
thus providing opportunity for both positive and negative directionality.
Section IV: Interpretation
For this Chi-Square analysis, this author will use the Pearson Chi-Square value and
significance value in Table 3 below. The Pearson Chi-Square value is 16.721. The p value = .000 which

is less than the significance level of = .05. Therefore, based upon the p value < .05, this author rejects
the null and accepts the alternative Ha which states there is a relationship between social problems and
dropout rates among the 9th grade student population.
It is important to note as well that the footnote below Table 3 indicates that 1 cell or 25% of the
Expected Counts was less than 5 with a lowest value of 1.14. This also confirms the lowest value data
provided in Table 2 indicating the assumption of expected frequencies was violated.
Table 3
Chi-Square Tests for independent variables SOCPROB and DROPOUT
Chi-Square Tests
Value
df
Asymp. Sig. (2-
Exact Sig. (2-
Exact Sig. (1-
sided)
sided)
sided)
Pearson Chi-Square
16.721a
.000
Continuity Correctionb
12.673
.000
Likelihood Ratio
11.305
.001
Fisher's Exact Test
Linear-by-Linear Association
N of Valid Cases
.001
16.531
.001
.000
88
a. 1 cells (25.0%) have expected count less than 5. The minimum expected count is 1.14.
b. Computed only for a 2x2 table
Note: Chi-Square Tests for independent variables SOCPROB and DROPOUT. Adapted from
The Phi and Cramers V test for independent variables SOCPROB and DROPOUT is provided
in Table 4 below. Although the Chi-Square test can detect whether there is an association between two
categorical independent variables, it cannot determine strength/effect of the relationship. Cramers V is a

measure of the strength association between two or more variables for any size of contingency tables,
especially large contingency tables. However, in a 2 x 2 contingency table like the one in this report,
Cramers V is equal to the absolute value of Phi () (see Table 4). If is between 0 and 1, it is
considered accurate in measuring the strength/effect of the association.
Phi () is calculated by dividing the Chi () value by the sample size (N) and then taking the
square root of the result. This was performed by PASW and provided in Table 4 below. One caution with
interpreting the value is when the sample size is considerably small; the value can become inflated.
When interpreting the value of , one uses the values .1, .3 and .5 as determinants to represent a small,
medium or large association respectively.
Table 4 below indicates = .436. Therefore this value falls between acceptable limits (0 <
< 1), indicating the measurement is accurate with a medium to large strength/effect of association
between the independent variables by comparing the value to .3 (medium effect) and .5 (large effect).
Table 4
Phi and Cramers V test for independent variables SOCPROB and DROPOUT
Symmetric Measures
Value
Nominal by Nominal
Approx. Sig.
Phi
.436
.000
Cramer's V
.436
.000
N of Valid Cases
88
Note: Phi and Cramers V test for independent variables SOCPROB and DROPOUT. Adapted from
This author will now refer again to the Cross Tabulation contingency table (Table 2) to further
interpret the percentage values of the association between the independent variables SOCPROB and
DROPOUT. This author has provided Table 2 again below for convenient comparisons and
interpretations. Under the SOCPROB variable No category, Table 2 below indicates 93.6% of 9th grade
students without social problems did not dropped out of school. By way of contrast, among the
remaining 6.4% of 9th grade students without social problems, only 50% dropped out of school.
Table 2
Cross Tabulation for variables SOCPROB and DROPOUT
SOCPROB * DROPOUT Cross Tabulation
DROPOUT
No
SOCPROB
No
Count
78
69.1
8.9
78.0
% within SOCPROB
93.6%
6.4%
100.0%
% within DROPOUT
93.6%
50.0%
88.6%
% of Total
83.0%
5.7%
88.6%
10
8.9
1.1
10.0
% within SOCPROB
50.0%
50.0%
100.0%
% within DROPOUT
6.4%
50.0%
11.4%
% of Total
5.7%
5.7%
11.4%
78
10
88
78.0
10.0
88.0
Count
Expected Count
Total
Total
73
Expected Count
Yes
Yes
Count
Expected Count
Note: The above is the Cross Tabulation table for SOCPROB and DROPOUT. Adapted from
Under the SOCPROB variable Yes category, Table 2 above indicates that 50% of 9th grade
students with social problems resulted in a 50% dropout percentage. By default, among the remaining
50% of 9th grade students with social problems, 50% did not drop out of school. Percentage participation
of 9th grade students is provided by adding the % of Total columns for both response categories (Yes
and No) which indicated a percentage participation of 100%.
To determine the Odds Ratio from the above Table 2, this author will look at 4 actual count
values. Among those who had no social problems (SOCPROB/No) and did not drop out
(DROPOUT/No), the actual count was 73 compared to 5 who had social problems (SOCPROB/Yes) and
did not drop out (DROPOUT/No). This indicates that 9th grade students without social problems are 14.6
(73/5) times less likely to drop out than those with social problems. Conversely, among those who had
no social problems (SOCPROB/No) and did drop out (DROPOUT/Yes), the actual count was 5 compared
to 5 who had social problems (SOCPROB/Yes) and also dropped out (DROPOUT/Yes). This indicates
that 9th grade students with or without social problems are equally (5/5) likely to drop out.
To determine the Odds Ratio this author will use the following formula: Odds no dropout /
Oddsdropout = 14.6/100 = 14.6%. This indicates that 9th grade students who did not have social problems
were 14.6% were more likely to not drop out of school. This is further illustrated in Figure 1 below
which indicates the no social problems had the lowest dropout rate. Figure 1 also helps verify the Cross
Tabulation contingency table (Table 2) which indicated that 93.6% of 9th grade students without social
problems did not drop out.

Figure 1: Bar Chart for independent variables SOCPROB and DROPOUT
Figure 1: The above is the Bar Chart for SOCPROB and DROPOUT. Adapted from ADDSC
scores [Data file], by Howell, David C., 2009, Statistical Home Page Web site, available from
http://www.uvm.edu/~dhowell/StatPages/
Section V: Conclusion and Limitations
Therefore, according to the Chi-Square test values provided in Table 3, the statistical notation
is as follows: 2 (1) = 16.721, p = .000 < .05, = .436. The measurement (0 < < 1) was determined
to be accurate with a medium to large effect size when comparing the value to .3 (medium effect) and .5
(large effect).
This author concludes through the Chi-Square test using an alpha level of .05, 1 degree of
freedom, = .436 and a p equal to .000 with a medium to large effect size of .436, that p is significant.
Therefore, H0 is rejected and Ha is accepted indicating there is a significant statistical association between
social problems and dropout rates among the 9th grade student population. This conclusion is further
substantiated by the Bar Chart diagram in Figure 1 of the independent variables SOCPROB and

DROPOUT. Moreover, because the variables were found to be dependent, the computed Odds Ratio
determined that 9th grade students were 14.6% more likely to not dropout whether with or without social
problems.
Limitations of the Chi-Square test above were: 1) As indicated in the Expected Counts (EC)
under both variables in Table 2, the ECs: 69.1, 8.9 and 8.9 were found to be greater than 5 and therefore
considered ideal. However, the EC in the Yes cell under variables SOCPROB and DROPOUT was
1.1 and less than 5 and cannot be considered ideal. Therefore the assumption of expected frequencies
was violated. This was further verified in the footnote below Table 3 which also indicated that 1 cell or
25% of the Expected Counts was less than 5 with a lowest value of 1.14. Secondly, it is important to
note that smaller sample sizes tend to inflate the Phi () value which indicated a medium to large
association.
References
Capella University. (2011). Inferential Statistics, Unit 2 Basic Concepts and Displaying Data,
Retrieved on August 20, 2011 from the Capella University, Minneapolis, Minnesota, Website:
http://courseroom2.capella.edu
Howell, D.C. (2008). Fundamental statistics for the behavioral sciences (6th ed.). Belmont, CA:
Thomson Wadsworth. ISBN: 9780495099000.
Howell, D.C. (2009). Statistical Home Page. ADD scores [Data file]. Retrieved July 24, 2011 from
http://www.uvm.edu/~dhowell/StatPages/
Howell, D. C., Huessy, H. R., & Hassuk, B. (1985). Fifteen year follow-up of a behavioral history of
attention deficit disorder. Pediatrics, 76, 185190.
Rhodes, Anthony. (2011). Capella University, Inferential Statistics, Unit Assignments, Retrieved on
September 4, 2011 from the Capella University, Minneapolis, Minnesota, Website:
http://courseroom2.capella.edu

Psy7620 U10a1 Chi Square

Uploaded by

Copyright:

Available Formats

Psy7620 U10a1 Chi Square

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Psy7620 U10a1 Chi Square

Uploaded by

Copyright:

Available Formats

What was the purpose of the chi-square test conducted?

What was the purpose of the chi-square test conducted?

What were the independent and dependent variables tested in the chi-square analysis?

What were the independent and dependent variables tested in the chi-square analysis?

Running head: u10a1 CHI SQUARE