The estimation procedure can run into numerical problems (usually in estimating standard errors) when variables are of dramatically different scale. Add a new Stata option RESCALE that gives the user the option to do the following:
Standardize all variables to mean zero and unit variance.
Estimate the model and standard errors with the standardized versions of the variables.
Rescale the model coefficients and standard errors to reverse the effects of the standardization.
Note that the standardization does not need to be exact since it will be reversed. The standardization does not affect the correct solution to the estimation problem, it only affects how well the algorithm does in finding the solution.
[x] Add RESCALE option, with scalar arguments "yes" and "no". The default is currently "no", but might be changed in the next major release.
The estimation procedure can run into numerical problems (usually in estimating standard errors) when variables are of dramatically different scale. Add a new Stata option RESCALE that gives the user the option to do the following:
Note that the standardization does not need to be exact since it will be reversed. The standardization does not affect the correct solution to the estimation problem, it only affects how well the algorithm does in finding the solution.