Summarize and ameans return no result within a loop

[Jose Federico Geli]

Hello Statalisters!

I need help to understand what I am doing wrong in my loop. If I run from the command line:

summarize epl_reg if year==2018 & country =="IE"

or

ameans epl_reg if year==2018 & country =="IE"

I am able to retrieve the scalar stored in r(mean).

However, if I run those commands within a loop, I eventually obtain no results (i.e. it reports 0 observations) from both the summary statistics or the ameans statistics. I say "eventually" because the first 5 or 6 times `l' takes the value "IE" and year takes the value 2018 both summary and ameans commands calculate the mean statistic correctly. The issue appears later. I do not understand why for a few times it does the proper calculation but then it does not.

The whole loop is the following:

        foreach y in $MacroList {
            foreach f in $EUFundsList {
                foreach x in $EUPaymentMethodList {
                    forvalues h = 1/$ForeHorizon {
                    
                    
                        display "fund=`f' y=`y' horizon=`h' instrumented_method=`x' instrument=`Inter_var'"
                        
                        qui sleep 250
                        preserve
                            
                        qui keep if fund == "`f'"                     
                        qui encode geo, gen(id_g)
                        qui gen geo_time=country+"_"+string(year)
                        qui encode geo_time, gen(id_ct)
                        qui xtset id_g year
            
                        
                        eststo clear
                            
                            local i=0
                            qui levelsof country, local(levels) 
                            foreach l of local levels {
                                
                                local i=`i'+1
                            
                                tempfile ft`i'
                                qui gen temp_var = `Inter_var'*`x'_cum_`h'
                                qui ivreghdfe     `y'_cum_`h' (c.`x'_cum_`h' c.temp_var = c.u_`x'_1 c.`Inter_var'#c.u_`x'_1 ) ///
                                l.`y'_cum_`h' c.`Inter_var' c.abs7, ///
                                absorb(id_g year) robust bw(`h') level(90) ///
                                partial(l.`y'_cum_`h' c.`Inter_var' c.abs7)
                                
                                display "country=`l'"
                                *summarize `Inter_var' if year==2018 & country =="`l'"        
                                ameans `Inter_var' if year==2018 & country =="`l'"
                                local Quant = r(mean)
                                
                                
                                qui parmby "lincomest `x'_cum_`h' +`Quant'*temp_var", label idstr("`l'") saving(`ft`i'',replace)
                                drop temp_var

                            }
                            
                            dsconcat `ft1' `ft2' `ft3' `ft4' `ft5' `ft6' `ft7' `ft8' `ft9' `ft10' `ft11' `ft12' `ft13' `ft14' `ft15' `ft16' `ft17' `ft18'
                            qui gen min90 = estimate - 1*stderr
                            qui gen max90 = estimate + 1*stderr  
                            
                            
                            qui gen macro = "`y'"
                            qui gen horizon = `h'
                            qui gen paymenttype= "`x'"
                            qui gen fund = "`f'"
                            
                            qui append using ResultsDBPanel_Marg_Int_`Inter_var'_Chart
                            qui save ResultsDBPanel_Marg_Int_`Inter_var'_Chart, replace                        

                            
                        restore
                        
                        */
                    
                    }    
                }
            }    
        }

My data looks like this for the year 2018 (790 observations, 3 for Ireland).

* Example generated by -dataex-. For more info, type help dataex
clear
input int year str3 country double epl_reg
2018 "AT" 2.2857143878936768
2018 "AT" 2.2857143878936768
2018 "AT" 2.2857143878936768
2018 "AT" 2.2857143878936768
2018 "AT" 2.2857143878936768
2018 "AT" 2.2857143878936768
....

2018 "HU" 1.5873016119003296
2018 "HU" 1.5873016119003296
2018 "IE" 1.2301586866378784
2018 "IE" 1.2301586866378784
2018 "IE" 1.2301586866378784
2018 "IT" 2.4742064476013184
2018 "IT" 2.4742064476013184
2018 "IT" 2.4742064476013184
....
end

Thank you very much!

Federico

[William Lisowski]

Perhaps the 0 observation results are for a country which does does not have any data in 2018

tab country year

will tell you if there are countries in your data which are not present in every year.

[Joro Kolev]

It is not clear to me what you are trying to achieve with this loop. The loop works as intended, but inside the loop you set a local

 
                                 *summarize `Inter_var' if year==2018 & country =="`l'"                                          ameans `Inter_var' if year==2018 & country =="`l'"                                  local Quant = r(mean)

and this local is overwritten each next round of the loop, and probably disappears anyways because it is local to the loop.

So what are you trying to achieve with your code in relation to your question?

You can also check out my egen function _gwmean, which is byable and can provide all types of weighted means.