Get predicted value using the coefficient

[Intan Salsabila Firman]

Hi, I have problem to calculate the predicted value of my variable, cashtoasset, for each observation. I use panel data with gvkey (cross sectional unit) and year (time unit). I use this model to calculate the predicted value for cashtoasset but i fail a lot when i use the loop function:

reg cashtoasset size markettobook cashflow ncworkingcapital bookleverage dividend acquisition


this is the problem i mean:

xtset gvkey year

Panel variable: gvkey (unbalanced)
Time variable: year, 1999 to 2022, but with gaps
Delta: 1 unit


. egen id=group(gvkey)

. gen predicted_cashtoasset=.
(6,302 missing values generated)

. qui sum id, meanonly

. local N= r(N)

. forvalues i=1(1)`N' {
reg cashtoasset size markettobook cashflow ncworkingcapital bookleverage dividend acquisition if id==`i'
predict px if id==`i'
replace predicted_cashtoasset = px if id==`i'
drop px
}
no observations
r(2000);

Do anyone know how calculate the predicted value ? Thank you

[Carlo Lazzaro]

Intan:
the issue seems to rest on the evidence that, for your first -gwkey- (panel identifier) , the variable -ncworkingcapital- is totally missing.
Therefore, -forval- (provided that a loop is necessary for what you're after) cannot move on and returns the error message.
That said, why, having a penel dataset, you -xtset- your data and then went -regress- instead of -xtreg-?

[Intan Salsabila Firman]

Ah thank you so much for reminding me about the number of missing observation of my variables and using xtreg. Because i saw it somewhere in the forum they suggest using reg for panel data in looping function instead of xtreg. I'm gonna change and try it once again.