logistic regression

[dpagliaccio]

Hi
Is there a way to set manymome to run logistic regression models glm(family=binomial)

[sfcheung]

Sorry for my late response. Officially, manymome does not support logistic regression. Some functions may work because the package just extracts the information (e.g., parameter estimates) from the output. The output of glm() is of the class glm, which is a subclass of lm. Therefore, many methods that work with lm objects also work with glm objects, though they may not know that an object is the output of glm().

So, using manymome with logistic regression output may be possible, but not officially supported. We try to be transparent in the process, so the computations of effects (indirect effects or conditional indirect effects) are reported by default. If you use it with logistic regression, always check the computation to see whether it makes sense.

Last, only the outcome variable (y variable) can be binary even in this case. Binary mediator needs special methods which is not supported in manymome.

We may formally add support for logistic regression but we do not yet have a schedule for it.

I will keep this issue open as we may add support for it in the future.