Modified Poisson adjusted risk ratio and confidence interval estimate

[yubaraj]

I have a binary outcome variable and covariates with multiple categories.

I am trying to estimate adjusted risk ratios and confidence intervals. 

I reviewed previous discussions in SAS  forum about this topic

https://communities.sas.com/t5/Statistical-Procedures/How-to-get-categorical-variables-RR-estimates-...

 https://communities.sas.com/t5/Statistical-Procedures/RR-estimates-using-proc-Genmod-for-categorical

https://academic.oup.com/aje/article/162/3/199/171116

345-2011: Using SAS® Procedures FREQ, GENMOD, LOGISTIC, and PHREG to Estimate Adjusted Relative Risk...

but these don't help when I have a covariate with multiple categories because the output it produces are not meaningful for multiple category.

It works well when I use lsmeans option

proc genmod data =xyz;
class id  x (ref = "5")  y (ref = "M")  z(ref = "Q1")  / param =glm;
model one_dose (event='1') = x y z/ dist = binomial  link = log;

lsmeans x y z / diff exp cl;

run;

but I want to use modified Poisson because prevalence of outcome is high. over 50% as a very similar study as mine used modified poisson regression.

https://www.thelancet.com/action/showPdf?pii=S2468-2667%2822%2900037-8

My code is as follows;

proc genmod data =xyz;
class id  x (ref = "5")  y (ref = "M")  z(ref = "Q1")  / param =ref;
model one_dose (event='1') = x y z/ dist = poisson link = log;
repeated subject = id/ type = unstr;
Estimate "RR x"  x  1/exp;
Estimate "RR y" y  1/exp;
Estimate "RR z" z 1/exp;
run;

So, basically, my question is how do we write the "estimate" statements when the covariate has multiple categories and how to assign a reference category in a modified Poisson model. 

Thanks a lot

Yuba 

[StatDave · Accepted solution]

Is there some reason that you feel you can't use the same LSMEANS statement in your second set of code like you did you your first set of code? See the Modified Poisson section of this note. 

[yubaraj]

Great @StatDave 

This modification works; I initially wrote param=ref, however, param=glm works as you mentioned in one of the  previous posts.

proc genmod data =xyz;
class id  x (ref = "5")  y (ref = "M")  z(ref = "Q1")  / param =glm;
model one_dose (event='1') = x y z/ dist = poisson link = log;
repeated subject = id/ type = unstr;

lsmeans x y z/ diff exp cl;

run;

Any idea how to assign a reference category, because it does not show results for first category as a reference category. 

[StatDave]

I don't know what you mean - LSMEANS gives an estimate for each level of the specified variable(s) and the DIFF option gives all of the pairwise differences. What and where exactly is the problem with respect to a reference level?