What does this mean? (multinomial logistic regression)

[arisoziro]

-Dataset file name; pa6

-Predictors; sedfive (0~4 categories) [meaning; sedentary behavior quantile, thus, getting sedentary from 0 to 4] (0;reference)

-dependent; htn (0~3 categories) [meaning; blood pressure normal (0;reference)-slightly high (1)-prehypertension(2)-hypertension(3)]

-I used the sas code below for above mentioned variables' relationships

proc logistic data=pa6;

class htn(ref="0") sedfive(ref="0") / param=ref;

model htn=sedfive / clodds=pl;

run; 

-then I got the output as attached.

-What does this output tell me? (How can I interpret it?)

-Do I have to use 'link=glogit' instead of 'clodds=pl'?

[StatDave]

Your response appears to be ordinal, so you probably do not want to use LINK=GLOGIT which would treat its levels as nominal (unordered). It is best to never specify the response in the CLASS statement. Instead use the response options immediately after the response variable in the MODEL statement. Whenever you have an ordinal response and you want to model the probability of higher response values, specify the DESCENDING response option (for example MODEL HTN(DESCENDING)=SEDFIVE; ). If you want to model the probability of lower response values, no response option is needed. ALWAYS look at the Response Profile table to see that the ordering of the levels is as desired. In your case, I suspect the Response Profile table shows that you are modeling the probability of lower response values. So, the odds ratio estimate labeled "sedfive 4 vs 0" (for example) means that the odds of a lower response value in the sedfive=4 level are 1.506 times the odds of a lower response value in the sedfive=0 level.