Regsave logistic regression

[Anne Todd]

Hello, using the wonderful regsave (https://ideas.repec.org/c/boc/bocode/s456964.html) to store results from regression models, it seems as if it still stores logit coefficients and CIs (as opposed to odds ratios) when using the -logistic- command.

Based on this: https://julianreif.com/regsave/ it seems like it has something to do with how Stata saves things in e(). Is there a way to transform that (and the CIs) within Stata before using regsave, perhaps?

[Daniel Schaefer]

Hi Anne, why do you need to store the odds ratios? Can't you just derive them later when you open the Stata dataset?

[Anne Todd]

[Deleted because of accidental double post]

[Anne Todd]

Sure, I could just write a function to derive them in Python/R later, but I was thinking if there's an easy way to do this in the process of saving with -regsave- that it would be simpler that way.

[William Lisowski]

Since the odds ratio = e^β you won't need to write a function to transform the coefficient estimates and confidence interval into odds ratios and confidence intervals.

Alternatively, the rtable option will solve your problem.

sysuse auto, clear
logit foreign mpg, or
regsave, ci
list

sysuse auto, clear
logit foreign mpg, or
regsave, rtable ci
list

. list, clean noobs

              var        coef     stderr    ci_lower    ci_upper    N  
      foreign:mpg    .1597621   .0525876    .0566922     .262832   74  
    foreign:_cons   -4.378866   1.211295   -6.752961   -2.004771   74

. list, clean noobs

              var       coef     stderr   ci_lower   ci_upper    N  
      foreign:mpg   1.173232   .0616975    1.05833   1.300608   74  
    foreign:_cons   .0125396   .0151891   .0011674   .1346911   74