PM.step() function

[mat194]

Hi everyone,

I have a question regarding the PM.step() function and its handling of categorical variables. Does the function account for the fact that some variables might be categorical and not purely numeric?

I attempted to replicate the results of the function to obtain coefficient estimates for my model implementation, but I noticed discrepancies between my results and those provided by the function. Additionally, it appears that the PM_data$new function does not accept non-numeric values for covariates. Am I missing something here?

For clarification, I am using the lm function to estimate the effects of my covariates on the parameters from the summarized data$cov results, not the entire dataset. Is this the correct approach? Should I be applying this to the EBE estimates instead?

Thank you very much for your assistance!

[mnneely]

Covariates must be numerical, otherwise, how would you incorporate them into equations? $step() is just a wrapper for the R step function. It uses the R lm function. Since all covariates are numbers, lm handles them as continuous, even though the values are discrete, instead of creating pseudovariables as it would for factors. This might affect the statistical results.