post-hoc for PROC FREQ

[Emma_at_SAS]

I have data that shows how many days during the pandemic, kids visited their friends. Overall, the number of days boys and girls visited their friends is significantly different. How may I test to see which interval categories are different as a post-hoc test? I am interested in 4 comparisons: girls 1-10d with boys 1-10d, girls 11-20d with boys 11-20d,  girls 21-30d with boys 21-30d, and girls 31+ with boys 31+

Thanks 

proc surveyfreq data =&data nosummary;
tables gender*visiting_friends_dpYear/ nostd nocellpercent row row(cl) chisq;
weight weight_var;
run;

The SURVEYFREQ Procedure

Gender	visiting_friends_dpYear	Frequency	Weighted Frequency		Row Percent	95% Confidence Limits for Row Percent
Girls	1-10	474	417.9		33.4	30.0	36.8
	11-20	239	210.0		16.8	14.1	19.5
	21-30	201	198.7		15.9	13.2	18.5
	31+	411	423.9		33.9	30.3	37.5
	Total	1325	1250.5		100.0
Boys	1-10	291	349.3		23.0	19.9	26.1
	11-20	248	354.9		23.4	20.1	26.6
	21-30	213	347.1		22.9	19.4	26.4
	31+	336	465.9		30.7	26.9	34.5
	Total	1088	1517.1		100.0
Total	1-10	765	767.1
	11-20	487	564.9
	21-30	414	545.7
	31+	747	889.8
	Total	2413	2767.6

Rao-Scott Chi-Square Test
Pearson Chi-Square	52.7567
Design Correction	1.7674
Rao-Scott Chi-Square	29.8491
DF	3
Pr > ChiSq	<.0001
F Value	9.9497
Num DF	3
Den DF	7236
Pr > F	<.0001
Sample Size = 2413

[Watts]

One possible approach is to construct the table as visiting_friends_dpYear * gender and use the DOMAIN=ROW option in PROC SURVEYFREQ to produce domain analysis (of gender) for each level of visiting_friends_dpYear. When you specify DOMAIN=ROW, the CHISQ option produces a one-way chi-square test for each level of the row variable (which compares 'girls 1-10d' with 'boys 1-10d' and so on).

proc surveyfreq data =&data nosummary;
tables visiting_friends_dpYear * gender / domain=row chisq;
weight weight_var;
run;