Bootstrap validation - Multivariable regression model

[Daniela Rodrigues]

Hello

I would like to perform bootstrap validation of a multivariable regression model, such as:

reg marks_end_year marks_beginning_year $student_characteristics i.discipline if year==1, base vce(robust)

How can I do this in Stata? I do not know if this has already been asked before. If so, could someone point me to those resources/ references?

Thanks in advance for your help

[Felix Bittmann]

It is very simple in Stata:

reg marks_end_year marks_beginning_year $student_characteristics i.discipline if year==1, base vce(bootstrap, reps(500))
estat bootstrap, bc

[Daniela Rodrigues]

Thanks very much Felix Bittmann. This worked. May I ask:

Would you be able to give some details about what is happening in each line of code?

Many thanks

[Felix Bittmann]

The first line is almost identical. Instead of computing robust standard errors we use bootstrapped ones with 500 replications. The larger this number, the more precise the result, on average. In the second line we compute bias-corrected CIs, which are probably better than the regular normal ones. Details are in the literature or in the stata helpfiles (help bootstrap) or online (https://www.stata.com/features/overv...nd-estimation/).

[Daniela Rodrigues]

Completely clear, thanks very much Felix Bittmann.