R 503 Conformability Issue | Help Please!!!

[Roman Johnson]

Hi, I am trying to estimate marginal effects after running a logistic model. The logistic model is fine but the post-estimation is giving me quite the headache. I am getting an error message saying that there is a conformability error. (I believe that means that there is a mismatch of the unit between my continuous income variable and the binary variable of trust.) I got this when running the following:

logit mammogramadherent i.race_r i.age i.education2 i.employment c.income i.family i.married i.citizenship i.obesity i.insurance i.
> provider i.safety i.trust i.placeofresidence i.produce i.racerxsafety i.racerxtrust i.racerxproduce


.
. mgen, at( income=(.5(5)60) trust=0) atmean stub(Some Trust) ///
> predlabel( Some Trust)
conformability error
r(503);

So, trying to troubleshot and fix the conformability error, I recoded the income continuous variable to a binary variable, e.g. income2.

I then received the following error message:


. estimates restore blm
(results blm are active now)

.
. mgen, at( income2=(.5(5)60) trust=0) atmean stub(Some Trust) ///
> predlabel( Some Trust)
at values for factor income2 do not sum to 1


Can someone please help me fix this conformability issue?

[Clyde Schechter]

-mgen- is not part of official Stata; it is community-contributed. Your post is the first time I've seen it mentioned on this Forum, which suggests that pehaps not many people use it. If nobody else responds in a timely manner, your best bet is to probably contact the author(s).

[Nick Cox]

I agree with Clyde Schechter

mgen is on SSC from Ben Jann who does post here.

Otherwise the issue over conformability is an issue over matrices in the same expression. Putting

set trace on

to see where the error arises and and then digging deeper to see the matrices in question might help.