I picked BLOOD/NATURAL FATHER EVER AN ALCOHOLIC OR PROBLEM DRINKER (S2DQ1 as the coding) as a candidate moderator. By common sense it’s not related with tobacco dependence. Let’s test it.

The source code is here, I hardcoded all unknown responses as missing. s3aq3b1 and tab12mdx are both category, but tab12mdx has only two levels, thus post-hoc test is not needed.

ods html close;
ods html;
libname worklib 'e:\Course\'; 
proc import
	datafile="E:\SPM\Coursera_datasets\nesarc_pds.csv"
	out=worklib.nesarc_pds;
	delimiter=',';
run;
data work;
	set worklib.nesarc_pds;
	select;
		when (s3aq3b1=9) s3aq3b1=.;
		otherwise;
	end;
run;
proc sort;
	by S2DQ1;
run;
proc freq;
	tables tab12mdx*s3aq3b1/chisq;
	by s2dq1;
run;

Here is the output. We should be more focused on the cases where BLOOD/NATURAL FATHER EVER AN ALCOHOLIC OR PROBLEM DRINKER is not missing. When S2DQ1=1, col pct is positively correlated between S3AQ3B1 and TAB12MDX=0, while S2DQ1=2, the same situation, both with p<0.0001. Thus, BLOOD/NATURAL FATHER EVER AN ALCOHOLIC OR PROBLEM DRINKER is not a moderator for this study.