Test of association tests whether or not any two categorical
variables are related to each other based on sample observations for instance: We may wish to test whether students’ choice of a degree program has anything to do with gender The two variables are ‘gender’ and ‘choice of degree program’ So we ask ourselves, “Are there differences in choice of degree program between male and female students?” The test will be able to tell you if choice of a degree program and gender are correlated but will not tell you directly what programs are most favored by men or by women for that matter
Tests for association
If the variables are related, we say that they are associated and therefore not independent Hypothesis testing is involved, you should follow the basic steps of hypothesis testing The tests are based on the Chi-squared tests model (classified under nonparametric statistical tests?) We seek to establish the Chi-squared test and critical statistics based on the sample evidence Tests for association Contingency table: Degree program Gender Economics Psychology Female 117 63 Male 24 56 Please take note that the two categorical variables under study are: Gender (either Male or Female) and Degree choice (Economics or Psychology) Is there an association between gender and the choice of degree program? Use a 5% level of significance to test for this.
Tests for association
H0: variables are not associated (independent) H1: variables are associated (not independent) The experimental data are presented in the form of a table called a contingency table A contingency table gives the frequency for two or more variables simultaneously It has rows and columns of observations for instance: A survey of first year university students sought to establish any association between choice of degree program and gender. Assuming only two degree programs were on offer, the following results were obtained: Tests for association-Suggested solution
Tests for association-Suggested solution
Tests for association-Suggested solution How do we compute expected frequencies (Ei)? Let us present the sample evidence with row and column totals: Degree program TOTALS Gender Economics Psychology Female 117 63 180 Male 24 56 80 TOTALS 141 119 260 Column, row and overall totals are as shown in the table
Tests for association-Suggested solution
Degree program TOTALS Gender Economics Psychology Female 117 63 180 Male 24 56 80 TOTALS 141 119 260 Tests for association-Suggested solution How do we compute expected frequencies (Ei)? Let us present the sample evidence with row and column totals: Degree program TOTALS Gender Economics Psychology Female 117 63 180 Male 24 56 80 TOTALS 141 119 260 Column, row and overall totals are as shown in the table
Tests for association-Suggested solution
Degree program TOTALS Gender Economics Psychology Female 117 63 180 Male 24 56 80 TOTALS 141 119 260 Tests for association-Suggested solution Degree program TOTALS Gender Economics Psychology Female 117 63 180 Male 24 56 80 TOTALS 141 119 260
Tests for association-Suggested solution
Degree program TOTALS Gender Economics Psychology Female 117 63 180 Male 24 56 80 TOTALS 141 119 260 Tests for association-Suggested solution Now we can present the table with both observed frequencies and their expected frequencies as follows: what was observed is not exactly as what was expected-there are deviations! Oi Ei 117 97.6154 63 24 43.3846 56 260 260
A national motor vehicle distributor wishes to find out if the size of the car bought is related to the age of the buyer. From the sales transactions over the past two years, a random sample of buyers were classified by size of car bought and buyer’s age. Assist this firm by performing a statistical test using α=0.01. Car size Buyer’s age Small Medium Large Under 30 years 10 22 34 30 to 45 24 42 48 Over 45 45 35 40