Skip an iteration within a loop

[Federico Nutarelli]

Hai all,

so I am performing bootstrap manually with a for loop. Basically what I am dong is to compute residuals from a first stage regression:

    capture quietly xtreg y L.major_recalls_norm average_age_prodbyatc3 avg_prd_sq mean_agefirm_byatc mean_agefirm_squared hhi share_expired share_patented i.Year, fe vce(cluster newid)
    capture predict residuals, e

and put them in a poisson fixed effects regression:

capture noisily quietly xtpoisson trials y residuals average_age_prodbyatc3 avg_prd_sq mean_agefirm_byatc mean_agefirm_squared hhi share_expired share_patented i.Year, fe vce(robust)

Now sometimes poisson fixed effects does not converge and an error like the following appears:

cannot compute an improvement -- discontinuous region encountered

which is my case at iteration 168. The code is:

set seed 1234
set matsize 11000
global filedata "DB_atc3.dta"
global varx "y residuals average_age_prodbyatc3 avg_prd_sq mean_agefirm_byatc mean_agefirm_squared hhi share_expired share_patented"
global resid "residuals"
global filename "bootexcel.xlsx"
global nrepl 300


local nx=0
set more off
foreach var in $varx {
local nx=`nx'+1 
}
global ncol= `nx' +2 

mat coeff = J($nrepl ,$ncol,.)

set more off
forval i=1/$nrepl {
use "$filedata", clear 
rename tot_count trials
*bysort idatc3 (Year): gen byte panelsize = _N
bsample, cluster(idatc3) idcluster(newid)
***
    capture xtset newid Year 
    capture quietly xtreg y L.major_recalls_norm average_age_prodbyatc3 avg_prd_sq mean_agefirm_byatc mean_agefirm_squared hhi share_expired share_patented i.Year, fe vce(cluster newid)
    capture predict residuals, e
    capture noisily quietly xtpoisson trials y residuals average_age_prodbyatc3 avg_prd_sq mean_agefirm_byatc mean_agefirm_squared hhi share_expired share_patented i.Year, fe vce(robust)
        
***
local conta=0
set more off
foreach var in $varx {
local conta=`conta'+1 
mat coeff[`i',`conta']=_b[`var']
}
mat coeff[`i',`conta'+1]=_b[$resid]

di as text " Replication # `i'"
mat coeff[`i',$ncol]=`i' 

}

What I would like to do is to ignore the error and keep looping. I wished I could do it with capture but it actually does not work. Is there a way to ignore iteration number 168 and carry on looping? Let's say I have 300 repetitions: the goal is to perform 299 repetitions ignoring the 168 (the one raising the error message)

Thank in advance,

Federico

[Bjarte Aagnes]

see

help continue

forvalues i = 1/10 {
    
    capture assert `i' != 6
    
    if ( _rc ) {
        
        continue
    } 
    
    di "`i'"
}

[Federico Nutarelli]

Bjarte Aagnes many thanks. I will check!

[Federico Nutarelli]

Bjarte Aagnes

EDIT: So sorry. I checked the help of capture which is quire new to me. I modified the code as follows but it does not quite work:

set seed 1234
forvalues i = 1/170 {
    use "/Users/federiconutarelli/Desktop/DB_atc3.dta", clear
    rename tot_count trials
    
    bsample, cluster(idatc3) idcluster(newid)
    
    capture assert `i' != 167 & `i' != 168
    
    if ( _rc ) {
        
              continue
    } 
    
    xtset newid Year 
    quietly xtreg y L.major_recalls_norm average_age_prodbyatc3 avg_prd_sq mean_agefirm_byatc mean_agefirm_squared hhi share_expired share_patented i.Year, fe vce(cluster newid)
    predict residuals, e
    quietly xtpoisson trials y residuals average_age_prodbyatc3 avg_prd_sq mean_agefirm_byatc mean_agefirm_squared hhi share_expired share_patented i.Year, fe vce(robust)
    di "`i'"
}

The problem I guess is more structural. Indeed I should ignore the error:

cannot compute an improvement -- discontinuous region encountered

Is there a way to ignore the error message and carry on the iteration ignoring it?

Thanks again!

[Bjarte Aagnes]

capture the command of interest, or possible a block of commands

capture xtpoisson 

if ( _rc ) {
    
    continue 
}

[Federico Nutarelli]

OH I see. Thanks a lot again!

[Necip Bulut]

Hi Bjarte Aagnes , I faced the same problem, could you please help me. My code is like that:

levelsof id, local(levels) // id is company id
foreach i of local levels{


nlsur (l_gm_Y=ln({A})+({s}/(1-{s}))*(ln(({alph}*gm_K^(({s}-1)/{s}))+((1-{alph})*((exp({g}))*gm_L))^(({s}-1)/{s})))) (l_KY=ln({alph})+(({s}-1)/{s})*(ln(gm_Y/gm_K)-ln({A}))) (l_LY=ln(1-{alph})+(({s}-1)/{s})*(ln(gm_Y/gm_L)-ln({A})-{g})), initial(aa), if id==`i'


replace coeff_A= _b[/A] if id==`i'
replace coeff_s= _b[/s] if id==`i'
replace coeff_alph= _b[/alph] if id==`i'
replace coeff_g= _b[/g] if id==`i'

}

My panel data include about 280 firms. I am trying to fit my equation for each firm and to save results. However, Stata does not solve the minimization problem for some firms such as firm 6. I have received error code r(430) and my loop does not continue for the rest of firms. I want that my loop codes ignore this firm and continue for other firms even if there is no solution for some firms. How can I do that?

[Bjarte Aagnes]

help capture

[EQ Teo]

Another way through which you can ignore the firms for which there is non-convergence when saving results in a loop is to add

  if e(converged) == 1 {
  
 local conta=`conta'+1  mat coeff[`i',`conta']=_b[`var']  
}

after the code for your regression.