Because I'm working on a logistic regression model with a large data set with many predictor variables, I decided to try working with speedglm.
When I execute
glm(y ~ x1+x2+x3+....+x100, family = "binomial", data = mydata)
it runs without issue, albeit with a warning that (at least for the full data set) that the model fails to converge (it works fine if I take a smaller set of predictor variables and/or a smaller data set). If I attempt the same as
speedglm(y ~ x1+x2+x3+....x100, family = "binomial", data = mydata)
I get the error
$ operator is invalid for atomic vectors.
mydata is a data frame, not a matrix of atomic vectors (hence glm has no problem with it). Why is speedglm returning this error?