Consider a dataset consisting of 610 males involved in a study of coronary heart disease.  The outcome variable is CHD status (1 = case, 0 = non case), the exposure variable of interest is CAT which is a dichotomous variable that indicates high (coded 1) or normal (coded 0) catecholamine level. The only other variables recorded in the data set are AGE (1 = age > 55, 0 = age ≤ 55) and ECG (1 = abnormal, 0 = normal). The dataset involving the above variables is given as follows: data data1; input cases total CAT AGE ECG; cards; 17 275 0 0 0 15 121 0 1 0 7 59 0 0 1 5 32 0 1 1 1 8 1 0 0 10 40 1 1 0 4 17 1 0 1 14 58 1 1 1 ; run; We are interested in the following logistic model:  As given in the SAS code above, the model is a full model with all main effects and interactions (both two way and three way interactions).  A main effect model can be obtained from SAS by this model statement –   model cases/total = AGE CAT ECG / cl;               Perform a hypothesis test to see if the interactions (including all of the two-way and the three-way Interactions) help with the model using a likelihood ratio test (LRT) to compare the full model and the main effect model using alpha of 0.05.

Consider a dataset consisting of 610 males involved in a study of coronary heart disease.  The outcome variable is CHD status (1 = case, 0 = non case), the exposure variable of interest is CAT which is a dichotomous variable that indicates high (coded 1) or normal (coded 0) catecholamine level. The only other variables recorded in the data set are AGE (1 = age > 55, 0 = age ≤ 55) and ECG (1 = abnormal, 0 = normal). The dataset involving the above variables is given as follows:

data data1;

input cases total CAT AGE ECG;

cards;

17 275 0 0 0

15 121 0 1 0

7 59 0 0 1

5 32 0 1 1

1 8 1 0 0

10 40 1 1 0

4 17 1 0 1

14 58 1 1 1

;

run;

We are interested in the following logistic model: 

As given in the SAS code above, the model is a full model with all main effects and interactions (both two way and three way interactions).  A main effect model can be obtained from SAS by this model statement –

model cases/total = AGE CAT ECG / cl;

            Perform a hypothesis test to see if the interactions (including all of the two-way and the three-way Interactions) help with the model using a likelihood ratio test (LRT) to compare the full model and the main effect model using alpha of 0.05.