09-27-2017 11:03 AM
I am running an analysis with a continuous outcome and a dichotomous outcome. Here's an outline of my table:
I was able to get a p-value for trend in both the bivariate analyses (using PROC GLM> estimate; and PROC FREQ> trend)
Now I need to determine if it is even possible to get p-value for trend in multivariate analyses. Theoretically, does that work? Here is my code:
**For continuous variable;
proc glm data = rongoana4 order = data; class hivpos3 (ref = '0') marital3 (ref = '1') lmpyear (ref = '2011') agecat4 (ref = 'Aged 15 or less') distcat (ref = 'Distance less than or equal to 5 km'); model p1v1_gest = hivpos3 marital3 lmpyear agecat4 distcat /solution; estimate "Linear trend for lmpyear" lmpyear -1 0 1; estimate "Linear trend for agecat4" agecat4 -0.564165 -0.298675 0.0995585 0.7632818; estimate "Linear trend for distcat" distcat -0.447043 -0.305405 -0.084097 0.8365453; run; quit;
**For dichotmous variable;
proc logistic data = rongoana4 desc;
class hivpos3 (ref = '0') marital3 (ref = '1') lmpyear (ref = '2011') agecat4 (ref = 'Aged 15 or less') distcat (ref = 'Distance less than or equal to 5 km')/param=ref;
model returnee = hivpos3 marital3 lmpyear agecat4 distcat / link=logit stb;*Removed gravida categories due to multicollinearity;
For the continuous variable, I don't know if my "estimate" requests are valid.
For the dichotmous variable, this guidance says that "In PROC LOGISTIC, the score test in the Testing Global Null Hypothesis: BETA=0 table is equivalent to the Cochran-Armitage test" so I am only getting one p-value. Is that the way it should be - just one global p-value?