Why is there an error message when I run a weighted logistic regression?

Asked May 12 '23 at 15:44

Active May 12 '23 at 15:44

Viewed 28 times

While performing a weighted logistic regression I got the error "notice! number of non-integer successes in a binomial glm" Why is this happening?

model<- glm(outcome~treatment, weights=weight, data=mydata, family=binomial(link=log))
betaiptw<- coef(model)
SE<- sqrt(diag(vcovHC(model, type ="HC0")))
rr<- exp(betaiptw[2])
lcl<- exp(betaiptw[2] -1.96*SE[2])
ucl<- exp(betaiptw[2] +1.96*SE[2])
c(lcl, rr, ucl)

I didn't expect the error message.

asked May 12 '23 at 15:44

ggplotr

1

That's the error you might get if `outcome` is not purely 1s and 0s – Allan Cameron May 12 '23 at 16:28
see e.g. `unique(mydata$outcome)`. – Axeman May 12 '23 at 16:36
The outcome is purely 1s and 0s. – ggplotr May 12 '23 at 17:49
In another thread, a user mentioned that the warning was just "glm being silly" https://stackoverflow.com/questions/50078497/weighted-logistic-regression-in-r. Can anyone confirm why this error is occurring? Should I be worried about it or nah? – ggplotr May 17 '23 at 15:26

Why is there an error message when I run a weighted logistic regression?

0 Answers0