A researcher has several variables that could be possible predictors for the final model. There is interest in checking all 2-way interactions for possible entry into the model. The researcher has dec

Question

proc logistic data=finances;
 class gender branch;
 model default = gender branch income debt
 gender|branch|income|debt @2
 / selection=forward slentry=0.01 <insert option here> ;
run;

Options

• Astart = 5
• Binclude = 4
• Cinclude = 5
• Dstart = 4

• A
  7% (2)
• B
  14% (4)
• C
  76% (22)
• D
  3% (1)

Explanation

In PROC LOGISTIC, the include=n option forces the first n effects into every model evaluated, and critically, the intercept counts as the first effect (position 1). With the intercept occupying slot 1 and the four main effects (gender, branch, income, debt) occupying slots 2–5, include=5 permanently forces all four main effects into the model, leaving only the six 2-way interactions as candidates for forward selection.

Why the distractors fail:

• B (include=4) forces only the intercept + 3 main effects, inadvertently leaving debt competing alongside the 2-way interactions-none of which would be guaranteed full consideration.
• A (start=5) and D (start=4) are wrong because start= is not a valid selection option in PROC LOGISTIC; it belongs to other SAS selection procedures such as PROC REG and PROC GLMSELECT.

Memory tip: In PROC LOGISTIC, mentally add 1 to the number of main effects you want to force: "4 main effects + 1 intercept = include=5." If you see start= in a PROC LOGISTIC answer choice, it's always a trap.

No community discussion yet for this question.