Response rate calculation from logistic regression using SAS

Hi All,

I am working on logistic regression. I have a binary response variable (PASI response:yes,no) and three categorical covariates-
treatment (treatment A, treatment B),
body weight (below 60kg, above or equal to 60kg)
and therapy taken (yes, no).

I need to calculate the following from the logistic regression model using SAS:

1. Adjusted response rate and standard error for treatment A and for treatment B,
2. Adjusted response rate difference and standard error between treatment A and treatment B,
3. 90% confidence interval for the adjusted response rate difference.

We use SAS in our organization for analysis. Can anyone please tell me how to do these using SAS?

Many thanks in advance !



Less is more. Stay pure. Stay poor.
Not following this, "Adjusted response rate and standard error for treatment A and for treatment B"

Do you mean the log odds or probability for A and B treatments when controlling for the other 2 covariates?
If so, you can get that from the model output, intercept and value for treatment variable. Then number two will be an estimate statement (perhaps with the diff option, and #3 you just need to change the value of alpha to 0.10.
Thanks for your reply. Attached is one of the requests I am working on. By adjusted response rate and standard error for treatment A , I mean the response rate for treatment A adjusted by the covariates in the model and std error of that estimate and similarly for treatment B. However, the request I am working on is not very clear to me. Before, I worked on calculating odds ratios, adjusted odds ratios and p-values for odds ratio etc fro logistic regression. But no idea how adjusted response rate comes from logistic model. Attached is the mock layout with title and footnote for the current request I am working:

Last edited:


Less is more. Stay pure. Stay poor.
Well your binary dependent variable is a response (yes/no) at 16 weeks. The logistic model would traditionally provide the log odds of response at 16 weeks, which can also be converted into the predicted probability, but this is not the response rate. Is this what is confusing you, because I not sure they know what they want or the label just needs to be changed from response rate.

Perhaps something with count data could be used, say negative binomial regression:
Not an area I am overly familiar with, so if you go that direction I would be interested in your approach.