Saving beta coefficients, constant and residuals from linear regression models? Is this code correct?

[Bjorn Sundstrom]

Hi, I am trying to save beta coefficients, constant and residuals and need advice if the following procedure returns the correct values.

Problem and cohort description: I am studying a disease outcome (lsf), over time (measure_disease_exam_duration) among a cohort of individuals (personal identification=pnr). I would like to make a regression model for each patient (pnr) and save the beta coefficient, constant and residual for each of the resulting model/patient. I have searched and tried some code examples, but have not been able to get them to work. I did manage to get the code below to work, but I am uncertain if the code returns the correct values.


*********************
*Stata 13.1

*Saving residual

sort pnr examdate

set more off

levelsof pnr, local(pnr)

gen residual_lsf = .

tempvar temp

foreach y of local pnr {

regress lsf measure_disease_exam_duration if pnr == `"`y'"'

predict `temp', resid

replace residual_lsf = `temp' if pnr == `"`y'"'

drop `temp'

}




*Saving coefficients and constants

sort pnr examdate

set more off

levelsof pnr, local(pnr)

gen lsf_beta = .

gen lsf_constant = .

tempvar beta1

foreach y of local pnr {

regress lsf measure_disease_exam_duration if pnr == `"`y'"'

replace lsf_constant = _b[_cons] if pnr == `"`y'"'

replace lsf_beta = _b[measure_disease_exam_duration] if pnr == `"`y'"'

}

[Nick Cox]

ssc install rangestat

local x  measure_disease_exam_duration
rangestat (reg) lsf `x',  int(examdate . .) by(pnr) 

gen residual = lsf - b_cons - b_`x' * `x'