cardiomoon / autoreg Goto Github PK

Hi,
the autoReg function throws an error when the model is of the form:
glm(y ~ x1 + x2, weights = w, data=data, family='binomial')

The error being:

Error in map():
ℹ In index: 1.
Caused by error in model.frame.default():
! 'data' must be a data.frame, environment, or list
Run rlang::last_trace() to see where the error occurred.

1. After digging deeper, the issue seems to stem from the function: fit2model, about line 77
   dataname = as.character(fit$call)[length(as.character(fit$call))]
   What happens is that, if you incorporate sample weights into the model, it appears as the last object in fit$call, and therefore dataname gets a vector of weights instead of the dataset used in fitting the model.

2. Another the issue seems to be coming from fit2list, as it fits a lm/glm model in about line 18, 21-22, with only, formula, data and family being parameters, there should probably be a ... so that extra parameters e.g weights are incorporated.

Thank you for the package. Packages creating publication ready tables are always welcome. I found that, at least for plain linear models, specifying new contrasts does not produce the final table as I expected. See the following for an example of what happens:

`library(autoReg)
data("mtcars")
fit=lm(mpg ~ cyl,
data=mtcars)
autoReg(fit,final=TRUE)

mtcars$cyl.f <- factor(mtcars$cyl)
fit = lm(mpg ~ cyl.f,
data=mtcars)
autoReg(fit,final=TRUE)

contrasts(mtcars$cyl.f) <- contr.treatment(3, base = 3)
fit = lm(mpg ~ cyl.f,
data=mtcars)
autoReg(fit,final=TRUE)`

Admittedly, I do not know how to read R code well enough to look at the underlying functions to know what is going on, or how it might be fixed.

Thanks again.