Re-scaling the second stage IV but using std of the first stage -- I need your assistance

[Khati Zolfaghari]

Hello, I use the "xtivreg2" analysis for my analysis, but because my instrument is a dummy variable I want to use the following steps to get the similar results to my baseline :

-I get " std" of my dummy variable
-"std" of the fitted value that I obtained from the first stage
-and then recalling the second stage by a ratio of these two standard deviations.

to get the std of my shock :

sum shock_recession

so to do this :

foreach v in "ltotalfertility" {            
foreach vart in "0"          {     

 xi: xtivreg2  d`vart'`v' (shock_recession = instrument) l(1/2).shock_recession l(1/1).d0`v' tt*, fe partial(tt*) robust  kernel(tru) bw(1) first  savefirst savefprefix(st1)

outreg2 using myfile,  replace label keep (shock_recession l(1/2)shock_recession l(1/1).d0`v') bdec(3) cdec(3) stats(coef se) addstat(kleibergen-Paap_rk_LM_statistic, `e(idstat)', p-value, `e(idp)', kleibergen-Paap_rk_LM_statistic, `e(widstat)') excel,


foreach vart in "1" "2" "3" "4" "5"            {     

 xi: xtivreg2  d`vart'`v' (shock_recession = instrument) l(1/2).shock_recession l(1/1).d0`v' tt*, fe partial(tt*) robust  kernel(tru) bw(1) first 

outreg2 using myfile,  append label keep (shock_recession l(1/2)shock_recession l(1/1).d0`v') bdec(3) cdec(3) stats(coef se) addstat(kleibergen-Paap_rk_LM_statistic, `e(idstat)', p-value, `e(idp)', kleibergen-Paap_rk_LM_statistic, `e(widstat)') excel,
}
}
}

but for getting the fitted value of first stage I tried in this way, because I need the standard deviation of first stage:

gen fitted=.
gen resi =.
foreach v in ltotalfertility { 
forvalues i=0/5 {



xi: xtivreg2  d`i'`v' (shock_recession = instrument) l(1/2).shock_recession l(1/1).d0`v' tt*,fe partial(tt*) robust  kernel(tru) bw(1) first 

predict f  if d`i'`v' 
    predict r if `i' `y', resid
    replace fitted= f if d`i'`v' & year==`i'
    replace resi = r  if d`i'`v' & year==`i'


}
}

but I get an error

predict not supported after xtivreg2 with partialling-out option
r(499);

and then

foreach v in ltotalfertility { 
    forvalues i=0/5{
    xi: xtivreg2  d`i'`v' (shock_recession = instrument) l(1/2).shock_recession l(1/1).d0`v' tt*,fe partial(tt*) robust  kernel(tru) bw(1) first 
    predict res_d`i'`v' if e(sample), res
    
}
}

error

option res not allowed
r(198);

how can I compute the "std" of the fitted value from the first stage? and re-scaling the second stage by the ratio of these two "standard deviations, std of my shock and std of the second stage.



I appreciate receiving your help. in any stage I receive errors.


regards,