Dataset from kaggle:

Link: https://www.kaggle.com/datasets/uciml/adult-census-income

Qualitative Variable:

Workclass – E.g., "Private", "Self-emp", "Government".

Marital Status – E.g., "Married", "Divorced", "Never-married".

Occupation – E.g., "Tech-support", "Craft-repair", "Sales".

Relationship – E.g., "Husband", "Wife", "Not-in-family".

Native Country – E.g., "United States", "Canada", "Mexico"

Ordinal Variable:

Education – Categories like "Preschool", "Some-college", "Bachelors", "Masters", "Doctorate" .

Income – The target variable: <=50K and >50K

Write description/reason of these tests and also write there formulas of Chi-square, Contingency
Coefficient, Phi & Cramer’s V, Lambda, Uncertainty Coefficient, Gamma, spearman, Kendal tau-b ,
Kendal tau-c, Somers’d:

1. Chi-Square (χ²) Test

Description:

 Measures the association between two categorical variables in a contingency table.

 Tests whether the observed frequencies differ significantly from the expected frequencies
under independence.
 Used in hypothesis testing for independence between categorical variables.

Formula:

χ2=∑ (O−E)²/E

Where:

 O= Observed frequency
 E = Expected frequency
 (calculated as Row Total×Column Total/Grand Total)

2. Contingency Coefficient (C)

Description:
  Measures the strength of association between two categorical variables.
 Derived from the chi-square test and ranges from 0 to 1 (higher values indicate stronger
association).
 Limited by the size of the contingency table.

Formula:

C= √ χ2/ χ2+N
Where:

 N= Total number of observations

 χ2= Chi-square value

3. Phi Coefficient (φ)

Description:

 Measures the association between two binary categorical variables (2×2 contingency
table).
 Similar to Pearson’s correlation coefficient but for categorical data.

Formula:

ϕ= √χ2/N
4. Cramer’s V

Description:

 A generalization of the Phi coefficient for larger contingency tables.

 Used when variables have more than two categories.

Formula:

V= √χ2/N(k−1)
Where:

 k = Minimum of (number of rows - 1) or (number of columns - 1)

5. Lambda (λ) Coefficient

Description:

 A measure of the proportional reduction in error (PRE) when predicting a dependent

categorical variable based on an independent categorical variable.
 Ranges from 0 (no association) to 1 (perfect prediction).

Formula:

λ=E1−E2/E1

Where:

 E1 = Number of errors made when ignoring the independent variable

 E2 = Number of errors made when using the independent variable

6. Uncertainty Coefficient (U)

Description:

 Measures the reduction in uncertainty of one variable given knowledge of another.

 Based on entropy and is useful for asymmetric associations.

Formula:

U=H(X)−H(X∣Y)/H(X)

Where:

 H(X) = Entropy of XXX

 H(X∣Y) = Conditional entropy of X given Y

7. Gamma (γ) CoefficientDescription:

 A measure of association for ordinal categorical variables.

 Accounts for concordant and discordant pairs.
 Useful for ranking relationships.

Formula:

γ=C−D/C+D
Where:

 C = Number of concordant pairs

 D = Number of discordant pairs

8. Spearman’s Rank Correlation (ρ)

Description:

 Measures the strength and direction of the association between two ranked variables.
 A non-parametric alternative to Pearson’s correlation.

Formula:

ρ=1−6∑di²/n(n²−1)

Where:

 di = Difference between ranks of corresponding values

 n = Number of pairs

9. Kendall’s Tau-b (τ_b)

Description:

 Measures association between two ordinal variables, adjusting for ties.

 Suitable for square tables.

Formula:

τb=C−D/ √ (C+D+Tx)(C+D+Ty)
Where:

 TxTy = Number of tied pairs in each variable

10. Kendall’s Tau-c (τ_c)

Description:

 A variant of Kendall’s Tau that accounts for different row and column sizes.
 Used when the number of categories in the two variables differs.
Formula:

τc=2(C−D)/n²(m−1)/m

Where:

 m = Minimum of the number of rows or columns

11. Somers’ D

Description:

 Measures the association between two ordinal variables, considering one as dependent.
 An asymmetric measure derived from Kendall’s Tau.

Formula:

DYX=C−D/C+D+TY
TY= Number of tied pairs in the dependent variable

1:Apply Chi-square test in SPSS:

Procedure: Analyze > Descriptive Statistics > Crosstabs.

Independent Variable: Education

Dependent Variable: Income

Independent Variable Dependent Variable Reason for Selection

Education (Ordinal) Income (Nominal: <=50K or Higher education may lead to
>50K) higher income

Crosstabs

[DataSet1] C:\Users\Almadina Computers\Desktop\Assignment1.sav

Case Processing Summary

Cases

Valid Missing Total

N Percent N Percent N Percent

education * income 820 82.1% 179 17.9% 999 100.0%

 education * income Crosstabulation

income Total

<-50k >50k

Count 59 98 157
Some-college
Expected Count 43.7 113.3 157.0

Count 17 110 127

Masters
Expected Count 35.3 91.7 127.0

Count 37 208 245

Education Bachelors
Expected Count 68.1 176.9 245.0

Count 1 48 49
Dectorate
Expected Count 13.6 35.4 49.0

Count 114 128 242

HS-grad
Expected Count 67.3 174.7 242.0
Count 228 592 820
Total
Expected Count 228.0 592.0 820.0

Chi-Square Tests

Value df Asymp. Sig. (2-

sided)

Pearson Chi-Square 101.441a 4 .000

Likelihood Ratio 109.130 4 .000
Linear-by-Linear Association 13.221 1 .000
N of Valid Cases 820

a. 0 cells (0.0%) have expected count less than 5. The minimum

expected count is 13.62.

Interpretation of Chi-Square Test Results

Significant Association: The Pearson Chi-Square test result (χ² = 101.441, p = 0.000) indicates a
statistically significant relationship between the two categorical variables tested.

Degrees of Freedom (df = 4): The test was conducted with four degrees of freedom, meaning
the variables had multiple categories.
Expected Counts Met Assumptions: No cells had an expected count less than 5, ensuring the
Chi-Square test is valid.

Linear Association: The Linear-by-Linear Association test (p = 0.000) suggests a strong

directional trend if the variables are ordinal.

Conclusion:

Since p < 0.05, we reject the null hypothesis, meaning the two variables are not independent—
they have a statistically significant relationship.

2:Contingency Coefficient:
Procedure: Analyze > Descriptive Statistics > Crosstabs.
One Nominal variable: Workclass
Another Nominal variable : Income Level
Independent Dependent Why
Workclass (Nominal) Income (Nominal) Does job type impact income?

Crosstabs

[DataSet1] C:\Users\Almadina Computers\Desktop\Assignment1.sav

Case Processing Summary

Cases

Valid Missing Total

N Percent N Percent N Percent

workclass * income 999 100.0% 0 0.0% 999 100.0%

workclass * income Crosstabulation

income Total

<-50k >50k

Workclass Count 16 14 30
?
Expected Count 8.6 21.4 30.0

Federal-gov Count 14 25 39

Expected Count 11.2 27.8 39.0

 Count 28 54 82
Local-gov
Expected Count 23.5 58.5 82.0

Count 185 436 621

Private
Expected Count 177.8 443.2 621.0

Count 6 67 73
Self-emp-inc
Expected Count 20.9 52.1 73.0

Count 32 84 116
Self-emp-not-inc
Expected Count 33.2 82.8 116.0

Count 5 33 38
State-gov
Expected Count 10.9 27.1 38.0
Count 286 713 999
Total
Expected Count 286.0 713.0 999.0

Symmetric Measures

Value Approx. Sig.

Nominal by Nominal Contingency Coefficient .173 .000

N of Valid Cases 999

Interpretation of the Contingency Coefficient in SPSS:

Contingency Coefficient (C) = 0.173, with a p-value = 0.000

Statistical Significance: Since p = 0.000 (< 0.05), we reject the null hypothesis, meaning there
is a statistically significant relationship between the two categorical variables.
Strength of Association:
C = 0.173 suggests a weak association between the two variables.
The closer C is to 1, the stronger the association; in this case, 0.173 is relatively low.

3:Phi & Cramér’s V:

Procedure: Analyze > Descriptive Statistics > Crosstabs.
One categorical variable : Occupation
Another categorical variable : Income Level
Statistics: Chi-Square
Phi & Cramer’s V

Independent Dependent Best Metric Reasons

Occupation (Nominal, Income (Nominal) Cramér’s V Different jobs have

multiple categories) different earnings
Crosstabs

Case Processing Summary

Cases

Valid Missing Total

N Percent N Percent N Percent

occupation * income 999 100.0% 0 0.0% 999 100.0%

occupation * income Crosstabulation

income Total

<-50k >50k

Occupation Count 16 14 30
?
Expected Count 8.6 21.4 30.0

Count 40 29 69
Adm-clerical
Expected Count 19.8 49.2 69.0

Count 0 1 1
Armed-Forces
Expected Count .3 .7 1.0

Count 50 69 119
Craft-repair
Expected Count 34.1 84.9 119.0

Count 22 207 229

Exec-managerial
Expected Count 65.6 163.4 229.0

Count 9 11 20
Farming-fishing
Expected Count 5.7 14.3 20.0

Count 8 7 15
Handlers-cleaners
Expected Count 4.3 10.7 15.0

Count 17 20 37
Machine-op-inspct
Expected Count 10.6 26.4 37.0

Count 17 7 24
Other-service
Expected Count 6.9 17.1 24.0

Count 37 194 231

Prof-specialty
Expected Count 66.1 164.9 231.0

Protective-serv Count 5 14 19

Expected Count 5.4 13.6 19.0

 Count 34 92 126
Sales
Expected Count 36.1 89.9 126.0

Count 8 27 35
Tech-support
Expected Count 10.0 25.0 35.0

Count 23 21 44
Transport-moving
Expected Count 12.6 31.4 44.0
Count 286 713 999
Total
Expected Count 286.0 713.0 999.0

Chi-Square Tests

Value df Asymp. Sig. (2-

sided)

Pearson Chi-Square 153.691a 13 .000

Likelihood Ratio 155.310 13 .000
N of Valid Cases 999

a. 3 cells (10.7%) have expected count less than 5. The minimum

expected count is .29.

Symmetric Measures

Value Approx. Sig.

Phi .392 .000

Nominal by Nominal
Cramer's V .392 .000
N of Valid Cases 999

Interpretation

Phi (.392, p = .000): Phi is a measure of association between two categorical variables. A value
of 0.392 suggests a moderate association. The p-value (.000) indicates that this association is
statistically significant.
Cramer's V (.392, p = .000): Cramer's V is similar to Phi but adjusts for table size. Since the
value is the same as Phi, it suggests a similar strength of association.
N of Valid Cases (999): This means 999 cases were used in the analysis.
4:Lambda:

Procedure: Analyze > Descriptive Statistics > Crosstabs.

Move one nominal variable: Workclass
Move another nominal variable: Income Level
 Click Statistics, then check:
✅ Chi-Square
✅ Lambda

Crosstabs

Case Processing Summary

Cases

Valid Missing Total

N Percent N Percent N Percent

workclass * income 999 100.0% 0 0.0% 999 100.0%

workclass * income Crosstabulation

Count

Income Total

<-50k >50k

? 16 14 30

Federal-gov 14 25 39

Local-gov 28 54 82

Workclass Private 185 436 621

Self-emp-inc 6 67 73

Self-emp-not-inc 32 84 116

State-gov 5 33 38
Total 286 713 999

Chi-Square Tests

Value Df Asymp. Sig. (2-

sided)
a
Pearson Chi-Square 30.997 6 .000
Likelihood Ratio 34.528 6 .000
N of Valid Cases 999

a. 0 cells (0.0%) have expected count less than 5. The minimum

expected count is 8.59.
 Directional Measures

Value Asymp. Approx. Approx. Sig.

a
Std. Error Tb

Symmetric .003 .008 .365 .715

workclass .000 .000 .c

Lambda Dependent

income .007 .019 .365 .715

Nominal by Nominal Dependent

workclass .003 .001 .003d

Dependent
Goodman and Kruskal tau
income .031 .009 .000d
Dependent

a. Not assuming the null hypothesis.

b. Using the asymptotic standard error assuming the null hypothesis.
c. Cannot be computed because the asymptotic standard error equals zero.
d. Based on chi-square approximation

Interpretation
Lambda Value Asymp. Sig. (p-value)
Symmetric (Overall λ) 0.003
Workclass Dependent 0.000
Income Dependent 0.007

Lambda (λ) = 0.003 (Symmetric) and 0.007 (Income Dependent):

These values are very close to 0, indicating that knowing one variable does not significantly
improve prediction of the other.

p-value (0.715 > 0.05):

Since the p-value is greater than 0.05, the association is not statistically significant.

Workclass Dependent (λ = 0.000):

This means that knowing "Workclass" does not help at all in predicting "Income Level".
5:Uncertainty Coefficient:
Procedure: Analyze > Descriptive Statistics > Crosstabs.
Move one nominal variable: Workclass
Move another nominal variable: Income Level
✅ Chi-Square
✅ Uncertainty Coefficient

Independent Dependent Why?

Workclass (Nominal) Income (Nominal) Does job type help predict
income?

Crosstabs

Case Processing Summary

Cases

Valid Missing Total

N Percent N Percent N Percent

workclass * income 999 100.0% 0 0.0% 999 100.0%

workclass * income Crosstabulation

Count

Income Total

<-50k >50k

? 16 14 30

Federal-gov 14 25 39

Local-gov 28 54 82

workclass Private 185 436 621

Self-emp-inc 6 67 73

Self-emp-not-inc 32 84 116

State-gov 5 33 38
Total 286 713 999

Chi-Square Tests
 Value Df Asymp. Sig. (2-
sided)
a
Pearson Chi-Square 30.997 6 .000
Likelihood Ratio 34.528 6 .000
N of Valid Cases 999

a. 0 cells (0.0%) have expected count less than 5. The minimum

expected count is 8.59.

Directional Measures

Value Asymp. Approx. Tb Approx. Sig.

Std. Errora

Symmetric .018 .006 3.206 .000c

workclass .013 .004 3.206 .000c

Nominal by Nominal Uncertainty Coefficient
Dependent

income Dependent .029 .009 3.206 .000c

a. Not assuming the null hypothesis.

b. Using the asymptotic standard error assuming the null hypothesis.
c. Likelihood ratio chi-square probability.

Interpretation of Uncertainty Coefficient in SPSS:

The relationship is statistically significant (p = 0.000), but the association is extremely weak (U < 0.03).
6:Gamma:
Procedure: Analyze > Descriptive Statistics > Crosstabs.
Move an ordinal variable: Education Level
Move another ordinal variable: Income Level
Click Statistics, then check:
✅ Chi-Square
✅ Gamma
Independent Dependent Why?
Education Level (Ordinal: Income Level (Ordinal: Low, Higher education may increase
Primary, High School, College, Medium, High) income.
etc.)

Crosstabs
 Case Processing Summary

Cases

Valid Missing Total

N Percent N Percent N Percent

education * income 820 82.1% 179 17.9% 999 100.0%

education * income Crosstabulation

Count

income Total

<-50k >50k
Some-college 59 98 157

Masters 17 110 127

education Bachelors 37 208 245

Dectorate 1 48 49

HS-grad 114 128 242

Total 228 592 820

Chi-Square Tests

Value df Asymp. Sig. (2-

sided)

Pearson Chi-Square 101.441a 4 .000

Likelihood Ratio 109.130 4 .000
Linear-by-Linear Association 13.221 1 .000
N of Valid Cases 820

a. 0 cells (0.0%) have expected count less than 5. The minimum

expected count is 13.62.

Symmetric Measures

Value Asymp. Std. Approx. Tb Approx. Sig.

a
Error

Ordinal by Ordinal Gamma -.194 .063 -3.064 .002

N of Valid Cases 820

a. Not assuming the null hypothesis.

 b. Using the asymptotic standard error assuming the null hypothesis.

Interpretation of Gamma in SPSS

1. Gamma (γ) = -0.194

This indicates a weak negative relationship between the two variables.

A negative Gamma means that as one variable increases, the other tends to decrease.

Since γ = -0.194, the relationship is weak (closer to 0).

2. p-value = 0.002

Since p < 0.05, the relationship is statistically significant.

This means the negative relationship is unlikely to be due to chance.

✅ The relationship is statistically significant (p = 0.002), but weakly negative (γ = -0.194).

7:Spearman:

Procedure: Analyze > Correlate > Bivariate.

Select two ordinal or continuous variables (e.g., Education Level and Income)
Under Correlation Coefficients, check:
✅ Spearman
Variable 1 (Ordinal) Variable 2 (Ordinal) Why?
Education Level (Ordinal: Income Level (Ordinal: Low, Higher education may lead to
Primary, High School, College, Medium, High) higher income.
etc.)

Nonparametric Correlations

Correlations

education Income

Correlation Coefficient 1.000 -.121**

Education Sig. (2-tailed) . .001

N 820 820
Spearman's rho
**
Correlation Coefficient -.121 1.000

Income Sig. (2-tailed) .001 .

N 820 999

**. Correlation is significant at the 0.01 level (2-tailed).

Interpretation:

 Correlation Coefficient (-.121):

The correlation between education and income is -0.121.

This is a weak negative correlation, meaning that as education increases, income

slightly tends to decrease. However, this is counterintuitive and might suggest
confounding factors or dataset issues.

 Significance (p-value = 0.001):

The p-value (Sig. 2-tailed) is 0.001, which is less than 0.01.

This means the correlation is statistically significant at the 1% level, so it's unlikely
to be due to random chance.

 Sample Size (N):

The number of observations (N) for education is 820, and for income, it is 999.

The difference in N suggests that some data might be missing for education when
compared to income.

The negative correlation is unexpected, as higher education is typically associated with

higher income.

8:Kendal tau-b:

Procedure: Analyze > Correlate > Bivariate.

Select two ordinal variables (e.g., Education Level and Income Level)
Under Correlation Coefficients, check:
✅ Kendall’s tau-b
Variable 1 (Ordinal) Variable 2 (Ordinal) Why?
Education Level (Ordinal: Income Level (Ordinal: Higher education may lead
Primary, High School, Low, Medium, High) to higher income.
College, etc.)

Nonparametric Correlations
 Correlations

education income

Correlation Coefficient 1.000 -.110**

education Sig. (2-tailed) . .001

N 820 820
Kendall's tau_b
**
Correlation Coefficient -.110 1.000

income Sig. (2-tailed) .001 .

N 820 999

**. Correlation is significant at the 0.01 level (2-tailed).

Interpretation of Kendall’s Tau-b in SPSS:

9:Kendall’s Tau-b (τb) = -0.110

This indicates a weak negative correlation between

education level and income.

A negative τb means that as education increases, income

tends to decrease, but the relationship is very weak.

Normally, education is expected to have a positive correlation

with income, so this result may suggest an inverse trend in the
dataset (e.g., overqualified individuals earning less in some
cases).

P-value = 0.001

Since p < 0.05, the relationship is statistically significant.

This means the weak negative correlation is not due to

random chance.

Sample Size (N = 820 for education, N = 999 for income)

The difference in sample sizes suggests there may be missing values in the
dataset for education.
This could affect the accuracy of the correlation

Conclusion

✅ There is a statistically significant but weak negative relationship

between Education Level and Income (τb = -0.110, p = 0.001).

10:Kendal tau-c:

Procedure: Analyze > Correlate > Bivariate.

Select two ordinal variables (e.g., Education Level and Income Level)
and move them to the Variables box.

 Under Correlation Coefficients, check:

✅ Kendall’s tau-c (Note: SPSS does not provide Tau-c directly, but it
can be computed using cross-tabulation).

Alternative Method (Using Crosstabs for Tau-c

Calculation)

1. Click Analyze > Descriptive Statistics > Crosstabs.

2. Move one ordinal variable to Row(s) (e.g., Education Level).
3. Move the other ordinal variable to Column(s) (e.g., Income Level).
4. Click Statistics, then check:
✅ Kendall’s tau-c

Variable 2 (Ordinal) Why?

Variable 1 (Ordinal)
Education Level Income Level Different category
(Ordinal: Primary, (Ordinal: Low, sizes make Tau-c
High School, Medium, High) more suitable.
College, etc.)

Crosstabs

Case Processing Summary

Cases

Valid Missing Total

N Percent N Percent N
 education * 820 82.1% 179 17.9% 999
income

education * income Crosstabulation

Count

Income Total

<-50k >50k

Some-college 59 98 157

Masters 17 110 127

education Bachelors 37 208 245

Dectorate 1 48 49

HS-grad 114 128 242

Total 228 592 820

Symmetric Measures

Value Asymp. Std. Approx. Tb

Errora

Ordinal by Ordinal Kendall's tau-c -.121 .040 -3.064

N of Valid Cases 820

a. Not assuming the null hypothesis.

b. Using the asymptotic standard error assuming the null hypothesis.

Interpretation of Kendall’s Tau-c in SPSS

Kendall’s Tau-c (τc) = -0.121

This indicates a weak negative correlation between

education level and income level.

A negative τc suggests that as education level increases, income level

slightly tends to decrease.

This contradicts the common expectation that higher education leads to

higher income, meaning there could be external factors affecting this trend
(e.g., job market saturation, industry variations, or underemployment).
P-value = 0.002 (< 0.05)

Since p < 0.05, the correlation is statistically significant.

This means that the observed weak negative relationship is

unlikely due to chance.

Standard Error = 0.040 & Test Statistic = -3.064

The test statistic (-3.064) suggests that the correlation is meaningful within
this dataset.

The standard error (0.040) indicates the reliability of the estimate; lower
values suggest a more precise measurement.

Conclusion

✅ There is a statistically significant but weak negative relationship

between Education Level and Income Level (τc = -0.121, p = 0.002).

11:Somers’d:

Procedure: Analyze > Descriptive Statistics > Crosstabs.

Move the independent ordinal variable (e.g., Education Level)
Move the dependent ordinal variable (e.g., Income Level)
 Click Statistics, then check:
✅ Somers' D

Dependent Variable Why?

Independent Variable (Outcome)
(Predictor)
Education Level Income Level (Low, Higher education
(Primary, High Medium, High) may lead to higher
School, College, etc.) income.

Interpret the SPSS Output:

Crosstabs
 Case Processing Summary

Cases

Valid Missing Total

N Percent N Percent N

education * 820 82.1% 179 17.9% 999

income

education * income Crosstabulation

Count

income Total

<-50k >50k

Some-college 59 98 157

Masters 17 110 127

education Bachelors 37 208 245

Dectorate 1 48 49

HS-grad 114 128 242

Total 228 592 820

Directional Measures

Value Asymp. Approx. Tb

Std. Errora

Symmetric -.104 .034 -3.064

education -.151 .049 -3.064

Ordinal by Ordinal Somers' d
Dependent

income Dependent -.080 .026 -3.064

a. Not assuming the null hypothesis.

b. Using the asymptotic standard error assuming the null hypothesis.

Interpretation of Somers' D in SPSS

Somers' D (Symmetric) = -0.104

This suggests a weak negative relationship between Education Level and Income Level.

The negative value means that as education level increases, income level tends to slightly
decrease.

This is unusual, as higher education is typically associated with higher income.

Somers’ D Interpretation
Dependent
Education Dependent (D = - Stronger negative effect Education is a weak
0.151) predictor of income, but in
this dataset, higher
education is associated with
slightly lower income.
Income Dependent (D = - Weaker negative effect Income is an even weaker
0.080) predictor of education level.
p-value = 0.002 (< 0.05)

The p-value is statistically significant, meaning the observed negative

relationship is unlikely due to random chance.

Standard Errors & Test Statistic (Z = -3.064)

The standard error is small, meaning the result is fairly reliable.

The test statistic (-3.064) supports the statistical significance of the

relationship.

Conclusion

✅ There is a statistically significant but weak negative relationship between

Education Level and Income Level (Somers' D = -0.104, p = 0.002).