running a loop n times with different outcome variables and storing the beta, standard error and p values

[Aditi Roy]

Hi, STATA people,

I have 12000 obs. I want to randomly assign fake treatment within each year with stipulated treatment and run the regression. I want to repeat this procedure 100 times. This is what my code is. I am unable to store the beta and standard values in the matrix format.

Also after

matrix p = J(100,`l',.)

STATA is issuing

invalid syntax
r(198);

Any help is appreciated. Struggling for a long time

set more off, permanently
set matsize 8000
set seed 8600
unab y : wheat lentils rice
local l : word count `y'
matrix p = J(100,`l',.)              
matrix colnames p =`y'
matrix b= J(100,`l',.)
matrix s= J(100,`l',.)
matrix colnames b =b `y'
matrix colnames s=se `y'
 forvalues i = 1/100 {
generate random=uniform()
gen byte treat= .
sort year random
by year: replace treat= (_n <=28) if (year == 1991)    //1991    28
by year: replace treat= (_n <=18) if (year == 1992)    //1992    18
by year: replace treat= (_n <=19) if (year == 1993)     //1993    19
by year: replace treat= (_n <=53) if (year == 1994)    //1994    53
by year: replace treat= (_n <=53) if (year == 1995)     //1995    53
by year: replace treat= (_n <=68) if (year == 1996)    //1996    68
by year: replace treat= (_n <=58) if (year == 1997)     //1997    58
by year: replace treat= (_n <=26) if (year == 1998)     //1998    26
by year: replace treat= (_n <=19) if (year == 1999)    //1999    19
by year: replace treat= (_n <=19) if (year == 2000)    //2000    19
by year: replace treat= (_n <=19) if (year == 2001)     //2001    19
by year: replace treat= (_n <=19) if (year == 2002)     //2002    19
by year: replace treat= (_n <=19) if (year == 2003)    //2003    19
by year: replace treat= (_n <=19) if (year == 2004)    //2004    19
by year: replace treat= (_n <=19) if (year == 2005)    //2005    19
by year: replace treat= (_n <=19) if (year == 2006)    //2006    19
by year: replace treat= (_n <=38) if (year == 2007)    //2007    38
by year: replace treat= (_n <=19) if (year == 2008)    //2008    19
by year: replace treat= (_n <=19) if (year == 2009)     //2009    19
by year: replace treat= (_n <=19) if (year == 2010)    //2010    19
by year: replace treat= (_n <=19) if (year == 2011)    //2011    19
by year: replace treat= (_n <=19) if (year == 2012)    //2012    19
by year: replace treat= (_n <=19) if (year == 2013)    //2013    19
by year: replace treat= (_n <=29) if (year == 2014)    //2014    29
by year: replace treat= (_n <=29) if (year == 2015)    //2015    29
by year: replace treat= (_n <=40) if (year == 2016)    //2016    40

foreach x of local i{
 local j = 1
foreach var of local y{
quietly reg `var' treat i.country_id  i.year i.country_id#c.line_time_trend [aw=ypop], cluster( country_id)
 matrix b[`i',`j++']= get(_b)
matrix s[`i',`j++']=get(_se)
matrix p[`i ',`j++'] = 2*ttail(e(df_r), abs(_b[treat]/_se[treat]))
 }
drop random
drop treat
// local ++i
}



// convert to dataset
// clear
svmat p, names(col)
svmat b', names(matnames)
svmat s, names(matnames)

[Phil Bromiley]

There is a procedure to automatically write the output of such estimations to a data set - I think Clyde has mentioned it on this listserve.
If you want these as data, why not
g b=.
g se=.
local a=0
foreach var .... (whatever) {
local a=`a'+1
regress y treat
replace b=_b[treat] in `a'/`a'
replace se=_se[treat] in `a'/`a'
}