Calcite | Level 5

## Is score data as count vs continuous

I have data from a pre-assessment and a post-assessment.  Each question/statement was coded as dichotomous (yes/no).  I took the sum of the questions for both pre (shown below) and post using an array:

array xst_itp [12] xst_1 xst_2 xst_3 xst_4 xst_5 xst_6 xst_7 xst_8 xst_9 xst_10 xst_11 xst_12;
pre_tot = sum(of xst_itp[*]);
pre_miss_n = cmiss(of xst_itp[*]);
if st_miss_n  = 12 then st_complete =  0;
if st_miss_n  < 12 then st_complete =  1;

I then ran a simple paired ttest:

proc ttest data = all_data;
paired pre_tot*post_tot;

run;

My question is, is it okay to treat this data as continuous data?  Do I need to treat it as count/discrete instead? If so, what analysis would you suggest in SAS?

My other question is, do I need to check the normality of both pre_tot and post_tot? And if one is not normal, do I need to use a Wilcoxon signed-rank test instead?

Super User

## Re: Is score data as count vs continuous

@lboyd wrote:

My question is, is it okay to treat this data as continuous data?  Do I need to treat it as count/discrete instead? If so, what analysis would you suggest in SAS?

That's more of a subject specific question in my opinion.

It depends on how you coded things and such. It's more counts really than a continuous distribution.

My other question is, do I need to check the normality of both pre_tot and post_tot? And if one is not normal, do I need to use a Wilcoxon signed-rank test instead?

You should check the normality, but I think the assumption still holds if the distributions are the same.

Discussion stats