Use this to substitute zero-values in your nrm values. Zero values in the predictors are recognized in the gHypEG regression as structural zeroes. To ensure this does not happen, please recode your zero-values in all your predictors, ideally using a dummy variable fitting an optimal value for the zeroes. This function takes a predictor that needs to be recoded and returns a list containing two matrices. The first one contains the original predictor recoded such that all zero values are 1 (and thus do not impact the model). The second one consist of a matrix with 1 where the original predictor was different from 0, and `zero_values` where the original predictor was 0. If `zero_values` is not specified, it is fixed to e to simplify the interpretation of the results.

get_zero_dummy(dat, name = NULL, zero_values = NULL)

Arguments

dat

matrix, the predictor for which the zero values should be recoded.

name

optional character, the name of the predictor to create a named list

zero_values

optional numeric, the value to assign to the zero values of `dat` in the dummy variable. It defaults to e to simplify the interpretation of the results.

Value

a possibly named list of two matrices. The first one is the recoded version of `dat` where all zeroes are changed to 1. The second is the dummy variable such that dummy[dat==0] <- zero_values and 1 otherwise.

See also