- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
I'm running the following multivariable Poisson model using PROC GENMOD, is there a way to change the units of the continuous variables baseline_SBP and baseline_DBP to 5 units instead of 1 unit? Similar to how UNITS is used in PROC LOGISTIC
proc genmod data = have;
class id gender(ref="M") race_cat(ref="White");
model bp_control = gender race_cat baseline_SBP baseline_DBP/ dist = poisson link = log;
repeated subject = id/ type = unstr;
run;
Accepted Solutions
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
I'm guessing that your response is binary and you are trying to estimate the relative risk as shown in the "Zou's modified Poisson approach" section this note - but for a 5 unit increase in those continuous variables in your case. Instead of the LSMEANS statement as shown in the note, you could use the ESTIMATE for your continuous variables. For example, use this statement to estimate the effect of a 5 unit increase. Be sure to look at the coefficients (shown by the E option) being applied to the parameter estimates to verify that you are estimating what you want.
estimate '5 unit RR' baseline_SBP 5 / exp e;
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
I'm guessing that your response is binary and you are trying to estimate the relative risk as shown in the "Zou's modified Poisson approach" section this note - but for a 5 unit increase in those continuous variables in your case. Instead of the LSMEANS statement as shown in the note, you could use the ESTIMATE for your continuous variables. For example, use this statement to estimate the effect of a 5 unit increase. Be sure to look at the coefficients (shown by the E option) being applied to the parameter estimates to verify that you are estimating what you want.
estimate '5 unit RR' baseline_SBP 5 / exp e;