Mod 9

Nonparametric
Tests
One & Two-Way Chi-Square Tests
Median Test
Mann-Whitney U Test
Kruskal-Wallis One-Way ANOVA

2
Key Concepts
*****
Nonparametric Tests
Parametric v nonparametric statistics

One-Way Chi-Square test and its assumptions
Observed and expected frequencies
Degrees of freedom in a One-Way Chi-Square test
Two-Way Chi-Square test
Expected frequencies and the rule of independent probabilities
Degrees of freedom in a Two-Way Chi-Square test
Chi-Square and the problem of small expected frequencies
Remedies for small expected frequencies
Yates’s Correction for Continuity
Fisher’s Exact Probability Test
Assumptions of the Two-Way Chi-Square test
The Median Test
Dealing with tied ranks in the Median Test
Degrees of freedom in the Median Test
Assumption of the Median Test
The Mann-Whitney U Test
The normal distribution and the Mann-Whitney U Test with large samples
Assumptions of the Mann-Whitney U Test
Dealing with tied ranks in a Kruskal-Wallis One-Way ANOVA
Assumptions of the Kruskal-Wallis One-Way ANOVA
Matrix of nonparametric statistical tests
Nonparametric Tests: Charles M. Friel Ph.D., Criminal Justice Center, Sam Houston State University
3
Lecture Outline
 Parametric v nonparametric statistics
 One-way Chi-Square Test
 Two-way Chi-Square Test
 The Median Test
 The Mann-Whitney U Test
 The Kruskal-Wallis one-way ANOVA
 An overview of various nonparametric

statistics
4
Nonparametric Statistics
Nonparametric Statistics A group of

statistical procedures that have two things in
common …
Designed to be used with nominal and/or

ordinal variables
Make few or no assumptions about

population parameters
Statistical Power All other things being equal,

nonparametric techniques are less powerful
tests of significance than their parametric
counterparts.
Assuming that the null hypothesis is false,
Nonparametric tests have less power to

discover significant relationships than
parametric tests.
5
One-Way Chi-Square Test
A test that compares the observed frequency of

a variable in a single group with what would be
the expected by chance.
Example Felony cases are assigned at random

to four district criminal courts. The annual
observed frequency of drug cases is given
below. Since the number of cases in each court
is not the same, is the case assignment system
not random?
District Court
132nd 189th 205th 264th
225 264 211 196
Total number of cases = 896
If the assignment process is truly random, we

would expect that (896 / 4) = 224 cases would
be assigned to each court by chance.
How much do the observed frequencies (fo)

differ from the expected frequencies (fe = 224)?
6
Calculating a One-Way Chi-Square
District Court
Frequencies 132nd 189th 205th 264th
Observed 225 264 211 196
Expected 224 224 224 224
χ 2 = Σ [ (fo – fe)2 / fe ]
χ 2 = (225 – 224)2 / 224 + (264 – 224)2 / 224 +

(211 – 224)2 / 224 + (196 – 224)2 / 224
χ 2 = (0.0045 + 7.1429 + 0.7545 + 3.50)
χ 2 = 11.40
7
Interpretation of Chi-Square
Chi-square is a family of probability distributions

that vary with degrees of freedom.
A one-way chi-square has df = (k – 1). For the

court example, df = (4 – 1) = 3 (k = # courts)
The critical value of χ 2

for 3 df at α = 0.05 is
7.815.
8
Since 11.40 > 7.815, the null hypothesis is

rejected and it is concluded that the assignment
system is not random.
9
An SPSS Example
Frequency of 70 Felony Cases

By Pretrial Status
In Jail On Bail ROR
32 21 17
If the null hypothesis assumes equal numbers of

cases in each status …
Do the observed frequencies differ

significantly from this assumption?
SPSS Results
Expected frequencies under
the null hypothesis.
NPar Tests
Chi-Square Test
Frequencies
PRE_STAT
Observed N Expected N Residual

1.00 32 23.3 8.7
2.00 21 23.3 -2.3
3.00 17 23.3 -6.3
Total 70
10
An SPSS Example One-Way Chi-Square Test (cont.)
Chi-square = 5.171,
df = (k - 1) = (3-1) = 2
p = 0.075
Test Statistics
PRE_STAT
Chi-Squarea 5.171
df 2
Asymp. Sig. .075
a. 0 cells (.0%) have expected frequencies less than
5. The minimum expected cell frequency is 23.3.
Interpretation
The distribution of cases by status is not

significantly different than the assumption of
equal numbers of cases in each pretrial
status category.
The differences between the observed and

expected frequencies are due to sampling
error.
11
Another SPSS Example

Assume the observed frequency of cases by

pretrial status given in the previous example.
Do they differ significantly from the expectation

that that there should have been 30 in jail, 20 on
bail, and 20 ROR?
SPSS Results
Expected frequencies
NPar Tests
Chi-Square Test
Frequencies
PRE_STAT
Observed N Expected N Residual

1.00 32 30.0 2.0
2.00 21 20.0 1.0
3.00 17 20.0 -3.0
Total 70
12
Another SPSS Example One-Way Chi-Square Test (cont.)
Chi-square = 0.633
df = 2
p = 0.729
Test Statistics
PRE_STAT
Chi-Squarea .633
df 2
Asymp. Sig. .729
a. 0 cells (.0%) have expected frequencies less than
5. The minimum expected cell frequency is 20.0.
Interpretation
The observed frequencies do not differ

significantly from the expected, and the
differences are due to sampling error.
13
Two-Way Chi-Square Test
In a one-way chi-square, the observed

frequencies for a single group are compared
with theoretically derived expected frequencies.
In a two-way chi-square, the observed

frequencies for two or more groups are
compared with expected frequencies derived
from the marginal totals of the cross-tabulation
table.
Example Is success on probation related to

drug addiction?
Not
Outcome Addicted Addicted Total
Success 115 277 392
Failure 194 57 251
Total 309 334 643
14
Given the marginal totals, what would the

expected frequencies be if drug addiction is
independent of success?
Calculating the Expected Frequencies
Under the multiplication rule of independent

probabilities, the probability of two events
occurring together is equal to their product.
Probability of success & addiction

P(S) = (392 / 643) = 0.6096
P(A) = (309 / 643) = 0.4813
P(S & A) = (0.6096) (0.4813) = 0.293
The expected number of successful
addicted probationers = (0.293) (643) = 188
By the same process, the expected frequencies

in the other cells are as follows:
S & NA = 204 F & A = 121
F & NA = 130
15
A simpler way to calculate the expected

frequencies
fe = (row total) (column total) / grand total
Observed & Expected Frequencies
Observed Frequencies
Not
Success 115 277 392
Failure 194 57 251
Total 309 334 643
Expected Frequencies
Not
Success 188 204 392

(0.2930) (0.3167)
Failure 121 130 251

(0.1876) (0.2028)
16
Total 309 334 643
17
Calculating Chi-Square
χ 2 = Σ [ (fo – fe)2 / fe ]
Calculation
χ 2 = (115 – 188)2 / 188 + (277 – 204)2 / 204 +

(194 – 121)2 / 121 + (57 – 130)2 / 130
χ 2 = 139.50
df = (r – 1) (c – 1) = (2 – 1) (2 – 1) = 1
Interpretation

for 1 df at α = 0.05
is 3.841.
Since 139.50 > 3.841, the null hypothesis is

rejected and it is concluded that success on
probation is related to addiction.
18
19
The Problem of Small Expected

Frequencies
The smaller the sample size, the greater the

likelihood that the expected frequencies in one
or more cells will be less than 5
When this happens, and depending upon the df

of the design, the chi-square distribution is not a
good fit to the data.
This may lead to a Type I error
Remedies
 Gather more data
 In designs larger than 2 x 2, collapse
categories with expected frequencies less
than 5
 In designs larger than 2 x 2, eliminate
categories with expected frequencies less
than 5
 In a 2 x 2 design, use Yates’s Correction
for Continuity
 In a 2 x 2 design, use Fisher’s Exact
20
Probability Test
Collapsing & Eliminating Categories

(Expected frequencies are in parentheses)
Single Married Divorced Widow Common Total

Law
Probation 8 (6.5) 5 (5) 4 (5) 0 (0.5) 3 (3) 20
Prison 5 (6.5) 5 (5) 6 (5) 1 (0.5) 3 (3) 20
Total 13 10 10 1 6 40
Collapsing Cells
Single Married Other Total
Probation 8 (6.5) 5 (5) 7 (8.5) 20
Prison 5 (6.5) 5 (5) 10 (8.5) 20
Total 13 10 17 40
Eliminating Cells
Single Married Divorced Total
Probation 8 (6.7) 5 (5.15) 4 (5.15) 17
Prison 5 (6.3) 5 (4.8) 6 (4.8) 16
21
Total 13 10 10 33
Yates’s Correction for Continuity

Yates’s correction involves reducing the
difference between (fo – fe) by 0.5 if fo is greater
than fe and increasing the difference by 0.5 if fo is
less than fe. This is accomplished as follows:
χ 2 = Σ [ (  fo – fe  - 0.5)2 / fe ]
Attitudes Towards the Death Penalty as a

Function of Gender (fe in parentheses)
Gender Approve Disapprove Total
Female 10 (13.26) 20 (16.74) 30
Male 9 (5.47) 4(7.26) 13
Total 19 24 43
χ 2
without Yates’s Correction
χ 2 = 4.76, df = 1, critical value of χ 2
at α = 0.5
equals 3.847
22
Conclusion Gender is significantly

associated with attitudes towards the death
penalty.
Yates’s Correction for Continuity (cont.)
χ 2
with Yates’s Correction
χ 2 = (  10 – 13.26  - 0.5)2 / 13.26 +

(  20 – 16.74  - 0.5)2 / 16.74 +
(  9 – 5.74  - 0.5)2 / 5.74 +
(  4 – 7.26  - 0.5)2 / 7.26
χ 2 = 3.41
Interpretation
χ 2 = 3.41, df = 1, critical value of χ 2

at α =
0.5 equals 3.847
23
Conclusion Gender is not significantly

associated with attitudes towards the death
penalty.
Type I Error
Without Yates’s correction, the analysis would

have resulted in a Type I error; rejection of the
null hypothesis when it is true.
24
Fisher’s Exact Probability Test
Used in a 2 x 2 design to calculate the exact

probability of the observed frequencies.
Death Penalty Convictions as a Function of Type of Counsel
Counsel Convicted Acquitted Total

A B
Appointed 6 2 8
C D
Retained 4 2 6
N
Total 10 4 14
P = [(A + B)! (C + D)!(A + C)!(B + D)!] / N!A!B!C!D!
P = [(6 + 2)! (4 + 2)!(6 + 4)!(2 + 2)!] / [14! 6! 2! 4! 2!]
P = [(8)! (6)!(10)!(4)!] / [14! 6! 2! 4! 2!]
P = (10!) / (14x13x …x9) (2!) (2!)
P = (3628800) / (2162160) (4) = 0.42
25
Interpretation Since 0.42 is > 0.05, the null

hypothesis is accepted; counsel is unrelated to
the disposition in death penalty cases.
26
An SPSS Example
Two-Way Chi-Square Test
Is there an association between race and the

type of sentence received by offenders
convicted of the same offence?
Race Type of Sentence Totals

Probation Prison
White 12 3 25
African American 12 10 22
Hispanic 13 10 23
Totals 37 33 70
SPSS Results
Crosstabs
Case Processing Summary
Cases
Valid Missing Total
N Percent N Percent N Percent
RACE * TYPE_SEN 70 100.0% 0 .0% 70 100.0%
27
An SPSS Example Two-Way Chi-Square Test (cont.)
RACE * TYPE_SEN Crosstabulation
Count
TYPE_SEN
.00 1.00 Total
RACE 1.00 12 13 25
2.00 12 10 22
3.00 13 10 23
Total 37 33 70
Chi-square = 0.386, df = 2, p = 0.825

Probability
Chi-Square Tests
Asymp. Sig.
Value df (2-sided)
Pearson Chi-Square .386a 2 .825
Likelihood Ratio .386 2 .825
Linear-by-Linear
.349 1 .555
Association
N of Valid Cases 70
a. 0 cells (.0%) have expected count less than 5. The
minimum expected count is 10.37.
Interpretation
The difference between the observed and

expected frequencies is not significant.
28
The observed differences among the racial

groups are due to sampling error.
29
The Median Test
Used to determine whether two random samples

come from populations with the same medians.
Assessment Center Rating by Two Teams:

Officer Candidates Randomly Assigned to
Assessment Teams
Team A Team B
Subject Rating Subject Rating
Goldin 72 Olsen 97
Jesani 67 Smither 76
Pritchard 87 Trantham 83
Birdwell 46 Gordon 69
Chavez 58 Graham 56
O’Neal 63 Andel 68
Johnson 84 Hutton 92
Tate 53 Paul 88
Bird 62 McGuire 74
Zuni 77 Costo 73
Compton 82 Raines 65
Lewis 89 Battan 54
Litzmann 43
30
The median test ( cont. )
Step 1 Determine the median rating for the two

assessment groups combined
Midpoint = (N + 1) / 2 = (25 + 1) / 2 = 13
Ranking of the combined ratings

of the two groups
97 69
92 68
89 67
88 65
87 63
84 62
83 58
82 56
77 54
76 53
74 46
73 43
72
median of combined groups
Step 2 Determine the number of officers in

either group whose ratings were equal to or
above the median, and the number not above
the median.
31
32
The median test (cont.)
Team A Team B
Above Mdn 6 7
Below Mdn 6 6
Step 3 Run a two-way chi-square to

determine whether there is an association
between assessment team and the ratings.
(expected frequencies are in parentheses)
Position Team A Team B Total
Above Mdn 6 (6.24) 7 (6.76) 13
Below Mdn 6 (5.76) 6 (6.24) 12
Total 12 13 25
χ 2 = (6 – 6.24)2 / 6.24 + (7 – 6.76)2 / 6.76 +

(6 – 5.76)2 / 5.76 + (6 – 6.24)2 / 6.24
33
χ 2 = 0.0367
34
The median test (cont.)
Interpretation

for 1 df at α = 0.05
is 3.841.
Since 0.0367 < 3.841, the null hypothesis is

accepted.
It is concluded that there is no

significant difference in the median
ratings given by the two assessment
teams.
35
An SPSS Example
The Median Test
Is gender related to the seriousness of the

offenses charged by the prosecutor?
Crime seriousness was graded on a 7-point

Likert scale where 1= very minor crime and 7 =
very serious crime.
SPSS Results
Codes for gender
0 = male
1 = female
Frequencies
GENDER
.00 1.00
SER_INDX > Median 15 11
<= Median 20 24
36
An SPSS Example The Median Test (cont.)
Chi-square = 0.979,
df = 1,
p = 0.322
Yates' corrected chi-square = 0.551

df = 1
P = 0.458
Test Statisticsa
SER_INDX
N 70
Median 4.0000
Chi-Square .979
df 1
Asymp. Sig. .322
Yates' Continuity Chi-Square .551
Correction df 1
Asymp. Sig. .458
a. Grouping Variable: GENDER
Interpretation
Male and female offenders do not differ

significantly in the median seriousness of the
crime with which they have been charged.
37
The Mann-Whitney U Test
Used to compare the ranks of two independent

groups, comparable to the purpose of the t test.
It is more powerful than the median test since

the latter only considers the number of cases
above & below the median, not the rank order of
the cases.
U = N1N2 + [N1(N1 + 1) / 2] - Σ R1
and
U′ = N1N2 + [ N2(N2 + 1) / 2 ] - Σ R2
Σ R1 = sum of the ranks for the smaller group
Σ R2 = sum of the ranks for the larger group
N1 = size of the smaller group
N2 = size of the larger group
38
Calculating a Mann-Whitney U
Assessment Center Rating By Two Teams:

Officers Randomly Assigned to Teams
Team A Team B
Score Rank (R1) Score Rank (R2)
72 13 97 25
67 10 76 16
87 21 83 19
46 2 69 12
58 6 56 5
63 8 68 11
84 20 92 24
53 3 88 22
62 7 74 15
77 17 73 14
82 18 65 9
89 23 54 4
43 1
Step 1 Rank the ratings from lowest to highest

regardless of assessment team.
39
Calculating a Mann-Whitney U (cont.)
Step 2 Sum the ranks in either group
Σ (R1) = 148
Σ (R2) = 175
Step 3 Calculate U
U = N1N2 + [N1(N1 + 1) / 2] - Σ R1
U = (12) (13) + [12 (12 + 1) / 2 ] - 148
U = 156 + 78 – 148 = 86
U′ = N1N2 + [N2(N2 + 1) / 2] - Σ R2
U′ = (12) (13) + [13(13 + 1) / 2] - 175
U′ = 156 + 91 – 175 = 72
40
Step 4 Determine the significance of U
Decide whether you are making a one- or a

two-tailed decision
Compare the smaller value of U to the

appropriate critical table value for N1 and N2
If the observed U is smaller than the table

value, the result is significant at the 0.05
level
Interpretation The critical value of U for N1 =

12 and N2 = 13, two-tailed α = 0.05, is 41.
Since the smaller obtained value of U (U′ = 72)

is larger than the table value,
The null hypothesis is accepted: there is no

difference in the ratings given by the two
assessment teams.
41
The Mann-Whitney U for Large Samples
U is normally distributed when
N1 > 20 and N2 > 40
In this case, the smaller value of U is

converted to a Z score, the critical values of
Z at 0.05 and 0.01 being 1.96 and 2.58
Z = [U – (N1N2) / 2 ] / N1N2(N1 + N2 + 1)/12
Suppose the assessment team data was based

upon samples of N1 = 30 and N2 = 35
Z = [ 72 – (30)(35)/2 ] / (30)(35)(30 + 35 + 1)/12
Z = -5.96
Interpretation
42
Since - 5.96 > 2.58, the results are

significant at p < 0.01. The assessment
teams differ in their ratings.
43
An SPSS Example
Mann-Whitney U Test
In a random sample of 70 offenders, were males

(0) charged with more serious crimes than
females (1)?
SPSS Results
NPar Tests
Mann-Whitney Test
Ranks
GENDER N Mean Rank Sum of Ranks

SER_INDX .00 35 39.03 1366.00
1.00 35 31.97 1119.00
Total 70
Mann-Whitney U = 489.00, p = 0.142
Z = -1.467, which is < the critical value of 1. 96, p > 0.05

Test Statisticsa
SER_INDX
Mann-Whitney U 489.000
Wilcoxon W 1119.000
Z -1.467
Asymp. Sig. (2-tailed) .142
a. Grouping Variable: GENDER
Interpretation
Gender is not related to crime seriousness
44
Kruskal-Wallis One-Way Analysis of

Variance
Used to compare three or more independent

samples with respect to an ordinal variable.
H = [ 12 / N (N+1) ] [ Σ ((Σ R)2 / n) ] – 3(N + 1)
N = the total number of cases
n = the number of cases in a given group
(Σ R)2 = the sum of the ranks squared for a

given
group of subjects
An Example
A state court administrator asked the 24 court

coordinators in the state’s three largest counties
to rate their relative need for training in case-
flow management on a Likert scale (1 to 7).
45
1 = no training need
7 = critical training need
Kruskal-Wallis One-Way Analysis of Variance (cont.)
Training Need of Court Coordinators
County A County B County C
3 7 4
1 6 2
3 5 5
1 7 1
5 3 6
4 1 7
4 6
2 4
4
5
Step 1 Rank order the total groups' Likert

scores from lowest to highest.
46
If tied scores are encountered, sum the tied

positions and divide by the number of tied
scores. Assign this rank to each of the tied
scores.
Kruskal-Wallis One-Way Analysis of Variance (cont.
Scores & Ranks Across the Three Counties
Ratings Ranks Ratings Ranks

1 2.5 4 12
1 2.5 4 12
1 2.5 5 16.5
1 2.5 5 16.5
2 5.5 5 16.5
2 5.5 5 16.5
3 8 6 20
3 8 6 20
3 8 6 20
4 12 7 23
4 12 7 23
4 12 7 23
Calculating the ranks of tied scores
47
Example Three court administrators rated

their need for training as a 3. These three
scores occupy the rank positions 7, 8, & 9.
(7 + 8 + 9) / 3 = 8
Kruskal-Wallis One-Way Analysis of Variance (cont.)
Step 2 Sum the ranks for each group and

square the sums
County A County B County C
Rating Rank Rating Rank Rating Rank

3 8 7 23 4 12
1 2.5 6 20 2 5.5
3 8 5 16.5 5 16.5
1 2.5 7 23 1 2.5
5 16.5 3 8 6 20
4 12 1 2.5 7 23
4 12 6 20
2 5.5 4 12
4 12
5 16.5
Σ R 67.0 153.5 79.5

4489 23562.25 6320.25
(Σ R)2
48
49
Kruskal-Wallis One-Way Analysis of Variance ( cont. )
Step 3 Calculate H
H = [ 12 / N (N+1) ] [ Σ ((Σ R)2 / n) ] – 3(N + 1)
H = [ 12 / 24 (24+1) ]
[4489 / 8 + 23562.25 / 10 + 6320.25 / 6]
– 3 (24 + 1)
H = (0.02) (3970.725) – (75)
H = 4.42
df = (k – 1) = (3 – 1) = 2
Interpretation
The critical chi-square table value of H for

α = 0.05, and df = 2, is 5.991
Since 4.42 < 5.991, the null hypothesis is

accepted. There is no difference in the
training needs of the court coordinators in
the three counties
50
An SPSS Example
Is race related to the seriousness of the offence

charged in a random sample of 70 offenders?
SPSS Results
White = 1, African American = 2, Hispanic = 3
NPar Tests
Kruskal-Wallis Test
Ranks
RACE N Mean Rank

SER_INDX 1.00 25 31.86
2.00 22 41.48
3.00 23 33.74
Total 70
Chi-square = 2.935, df = 2, p = 0.23

Test Statisticsa,b
SER_INDX
Chi-Square 2.935
df 2
Asymp. Sig. .230
a. Kruskal Wallis Test
b. Grouping Variable: RACE
Interpretation
No significant relationship found between

race and crime seriousness.
51
52
Matrix of Nonparametric Statistics
Level of Nonparametric Test Nonparametric

Measurem Correlation
ent
One Sample Two Samples K Samples
Related Unrelated Related Unrelated
Nominal Binomial McNemar Fisher Cochran Chi- Cramer Coefficient

Test Change Exact Test Q Test Square
Test for 2x2 Test for Phi Coefficient
Chi-Square Tables rxk Tables
Test Kappa Coefficient
Chi-
Square Asymmetrical Lambda
Test for Statistic
rx2 Tables
Ordinal Kolmogorov- Sign Test Median Friedman Extension Spearman Rank-Order

Smirnov Test Two- Way of the Coefficient
One- Wilcoxon ANOVA Median
Sample Test Signed Mann- by Ranks Test Kendall Rank-Order
Ranks Whitney U Coefficient
One-Sample Test Test Page Test Kruskal-
Runs Test for Ordered Wallis Kendall Partial Rank-
Robust Alternatives One-Way Order Coefficient
Change- Rank- ANOVA
Point Test Order Test Kendall Coefficient of
Jonckheere Concordance
Kolmogoro Test for
v- Ordered Kendall Coefficient of
Alternatives
Smirnov Agreement
Two-
Sample Correlation Between k
Test Judges and a Criterion
Test
Siegel-
Tukey Test Gamma Statistic
for Scale
Differences
Somer’s Index of
Asymmetric
Association

Mod 9

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

Mod 9

Uploaded by

Copyright:

Available Formats

Nonparametric

Kruskal-Wallis One-Way ANOVA

Parametric v nonparametric statistics

 Parametric v nonparametric statistics

 One-way Chi-Square Test

 Two-way Chi-Square Test

 The Median Test

 The Mann-Whitney U Test

 The Kruskal-Wallis one-way ANOVA

 An overview of various nonparametric

Nonparametric Statistics A group of

Designed to be used with nominal and/or

Make few or no assumptions about

Statistical Power All other things being equal,

Assuming that the null hypothesis is false,

Nonparametric tests have less power to

One-Way Chi-Square Test

A test that compares the observed frequency of

Example Felony cases are assigned at random

225 264 211 196

Total number of cases = 896

If the assignment process is truly random, we

How much do the observed frequencies (fo)

Calculating a One-Way Chi-Square

Frequencies 132nd 189th 205th 264th

Observed 225 264 211 196

Expected 224 224 224 224

χ 2 = (225 – 224)2 / 224 + (264 – 224)2 / 224 +

χ 2 = (0.0045 + 7.1429 + 0.7545 + 3.50)

Chi-square is a family of probability distributions

A one-way chi-square has df = (k – 1). For the

The critical value of χ 2

Since 11.40 > 7.815, the null hypothesis is

Frequency of 70 Felony Cases

In Jail On Bail ROR

If the null hypothesis assumes equal numbers of

Do the observed frequencies differ

Observed N Expected N Residual

An SPSS Example One-Way Chi-Square Test (cont.)

The distribution of cases by status is not

The differences between the observed and

Another SPSS Example

Assume the observed frequency of cases by

Do they differ significantly from the expectation

Observed N Expected N Residual

Another SPSS Example One-Way Chi-Square Test (cont.)

The observed frequencies do not differ

Two-Way Chi-Square Test

In a one-way chi-square, the observed

In a two-way chi-square, the observed

Example Is success on probation related to

Success 115 277 392

Failure 194 57 251

Total 309 334 643

Given the marginal totals, what would the

Calculating the Expected Frequencies

Under the multiplication rule of independent

Probability of success & addiction

By the same process, the expected frequencies

A simpler way to calculate the expected

fe = (row total) (column total) / grand total

Observed & Expected Frequencies