foreach issues

[Andrea Stringhetti]

Hi everyone! I have some issues in running two loops.
What I want to do is to estimate alpha and beta coefficients for each id and date in my dataset. what I want to obtain is a different alpha and beta and residuals for each line.
What I implemented is the following code:

g alpha=. 
g beta1=.
g residuals=.
​​​​​​​
qui levelsof(date), local(date)
foreach date of local date {

    levelsof id, local(levels) 
    foreach v of local levels  {
    
    gen date_250 = date-250 if date == `date'
    gen date_50 = date-50  if date == `date'
    egen d250 = mean(date_250)
    egen d50 = mean(date_50)
    format %tdDD/NN/CCYY date_250
    format %tdDD/NN/CCYY date_50
    format %tdDD/NN/CCYY d250
    format %tdDD/NN/CCYY d50
    gen check = (date>=d250 & date<=d50)

    quietly reg lnstock_return lnmkt_return if id ==`v' & check ==1, r
    predict temp2 if id== `v', resid 
    replace residuals = temp2 if id == `v' & date == `date'
    drop temp2
    replace alpha = _b[_cons] if id ==`v' & date == `date' 
    replace beta1 = _b[lnmkt_return] if id ==`v' & date == `date'
    }

drop check d50 d250 date_*
}

However, my loops are not running correctly and I have an estimate only for one alpha and one beta, without any error message.

Do you have any idea of why?

Thanks for your help,

Andrea

[Sven-Kristjan Bormann]

You should definitely see an error message which should tell you that you cannot generate a variable more than once. You need to move the line

drop check d50 d250 date_*

to the end of

foreach v of local levels  {
...
}

or you move the lines

   gen date_250 = date-250 if date == `date'
    gen date_50 = date-50  if date == `date'
    egen d250 = mean(date_250)
    egen d50 = mean(date_50)
    format %tdDD/NN/CCYY date_250
    format %tdDD/NN/CCYY date_50
    format %tdDD/NN/CCYY d250
    format %tdDD/NN/CCYY d50
    gen check = (date>=d250 & date<=d50)

into the first loop, for example before the line

levelsof id, local(levels)

.

[Andrea Stringhetti]

Thank you very much! I followed your advice but aplha, beta1 and residuals variables are still missing after the loop ended.

[cu dao huy]

Hi Andrea,

I think you should name your local variables in different way. For example, your codes:

qui levelsof(date), local(date)
foreach date of local date {
    levelsof id, local(levels)
    foreach v of local levels  {
         gen date_250 = date-250 if date == `date'
         gen date_50   = date-50   if date == `date'

I dont know which `date' Stata will pickup in line

gen date_250 = date-250 if date == `date'

. The `date' in levelof or in foreach?

you could try "set trace on" command before your loop to see how it work and locate the problem!

[Sven-Kristjan Bormann]

Stata will set the macro to empty if it sees a construct like

foreach date of local date {

This should explain why the code does not work. Unfortunately, Stata does not complain if it sees such a construct and will happily set the macro to empty.

[cu dao huy]

Stata will set the macro to empty if it sees a construct like

foreach date of local date {

This should explain why the code does not work. Unfortunately, Stata does not complain if it sees such a construct and will happily set the macro to empty.

Hi, pls put it in the wishlist for Stata 17 :>

[Sven-Kristjan Bormann]

Hi, pls put it in the wishlist for Stata 17 :>

I have only Stata 14. Do know you if the same behavior is observed also in Stata 16? Before I put it on the wishlist for Stata 17, I want to be sure that this behavior has not been fixed yet.

[cu dao huy]

I have only Stata 14. Do know you if the same behavior is observed also in Stata 16? Before I put it on the wishlist for Stata 17, I want to be sure that this behavior has not been fixed yet.

Hi, I am using Stata version 15.1 SE window (64-bit x86-64). And trying:

local date 12345 12345 12345
foreach date of local date {
    di `date'
}

It's still happen on Stata 15.1 version! I think Stata 16 is still as well!

[Sven-Kristjan Bormann]

Ok. I have added it to the wishlist.