qregsel implementation

[Co Ar]

Dear members
Am implementing the recommendations in this paper https://www.stata.com/meeting/us20/s..._Siravegna.pdf
My selection variable is sex ie select(sex) coded 0, 1.
However, when I run the model as follows:

qregsel yield age educ , select( sex ) quantile(.1 .5 .9) grid_min(-.9) grid_max(.9) grid_length(0.05)

This error keeps recurring

__000001 should be binary.
r(198);
What I doing wrong please

[Stephen Jenkins]

It would help if you reassure us and yourself that "sex" really is binary. Show us the results of

tabulate sex
tabulate sex, missing

Do you get the same error if you omit the space before and the space after 'sex' in the select() option?

What if you run the command adding an -if- qualifier of form -if inlist(sex,0,1)-?

With a log-file open, -set trace on- and run your command. -set trace off- at the end. Look at the log-file. Does the detailed output around the error provide more clues about the problem?

[Co Ar]

I have re-run the diagnostics:
for the

 
 tabulate sex

I get:

sex	Freq.	Percent	Cum.
0	131	15.07	15.07
1	738	84.93	100.00
Total	869	100.00

For next

tabulate sex, missing

I get

sex	Freq.	Percent	Cum.
0	131	15.07	15.07
1	738	84.93	100.00
Total	869	100.00

For trace:
I get

- 
- syntax varlist(numeric) [if] [in], SELect(string) quantile(string) grid_min(real) grid_max(real) grid_length(real) [ copula(string) noCONStant plot ]
- gettoken depvar indepvars : varlist
- _fv_check_depvar `depvar'
= _fv_check_depvar yield
- fvexpand `indepvars'
= fvexpand  age educ
- local cnames `r(varlist)'
= local cnames age educ
- tokenize `select', parse("=")
= tokenize sex, parse("=")
- if "`2'" != "=" {
= if "" != "=" {
- local x_s `select'
= local x_s sex
- tempvar y_s
- qui gen `y_s' = (`depvar'!=.)
= qui gen __000001 = (yield!=.)
- }
- else {
  local y_s `1'
  local x_s `3'
  }
- capture unab x_s : `x_s'
= capture unab x_s : sex
  --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- begin unab ---
  - version 6
  - gettoken user 0: 0, parse(" :")
  - gettoken colon 0: 0, parse(" :")
  - if `"`colon'"' != ":" { error 198 }
  = if `":"' != ":" { error 198 }
  - syntax [varlist(default=empty)] [, MIN(integer 1) MAX(integer 120000) NAME(string)]
  - c_local `user' `varlist'
  = c_local x_s sex
  - local n : word count `varlist'
  = local n : word count sex
  - if `n'>=`min' & `n'<=`max' { exit }
  = if 1>=1 & 1<=120000 { exit }
  ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- end unab ---
- marksample touse
- markout `touse' `y_s' `x_s'
= markout __000002 __000001 sex
- tempvar touse1
- mark `touse1' if `y_s'==1
= mark __000003 if __000001==1
- markout `touse1' `depvar' `indepvars'
= markout __000003 yield  age educ
- qui replace `touse' = 0 if `touse1'==0 & `y_s'==1
= qui replace __000002 = 0 if __000003==0 & __000001==1
- qui tab `y_s'
= qui tab __000001
- if r(r) != 2 {
- dis as error "`y_s' should be binary."
= dis as error "__000001 should be binary."
__000001 should be binary.
- exit 198
  }

[Co Ar]

Below is part of the data

* Example generated by -dataex-. To install: ssc install dataex
clear
input float yield byte(sex age educ)
     1000 1 28 2
      240 0 60 2
      250 0 40 4
      125 1 41 2
      125 1 54 2
        4 1 72 2
 666.6667 1 35 2
 352.9412 1 53 2
      225 0 48 2
      200 1 56 2
      200 1 42 2
293.33334 1 59 2
       96 1 81 2
       10 0 41 2
       10 1 39 2
      160 1 45 2
      600 1 33 2
       65 1 47 2
       72 0 60 1
      255 0 44 2
       50 1 38 2
       42 0 49 2
      120 1 38 2
       30 1 35 2
       50 0 53 2
       40 1 52 2
      480 0 68 2
       40 1 50 2
13.333333 1 57 2
      700 1 63 2
       60 1 47 2
      180 1 41 2
      240 1 32 2
      126 1 36 2
      180 1 42 2
    56.25 1 58 2
       60 1 29 1
      150 1 37 2
      420 1 41 2
      760 1 43 2
      110 1 34 2
       90 1 47 2
 8.333333 1 67 1
      100 1 40 3
      180 1 50 1
     37.5 1 47 2
      148 0 42 2
      150 1 75 1
 362.6667 1 47 2
      455 1 29 2
      420 1 45 2
      150 1 65 2
       30 1 40 2
       30 1 65 2
      150 1 30 2
       30 1 69 1
       30 1 25 3
       75 1 58 1
      120 1 70 1
      390 1 65 2
      140 1 36 1
       50 1 47 1
      300 1 25 2
      500 1 47 1
       10 1 27 2
      200 0 32 1
       84 0 44 2
      210 0 33 2
       15 1 52 2
       72 0 60 4
      192 0 39 3
      100 0 68 2
       75 0 40 2
      120 0 43 2
      200 1 51 2
       80 0 42 2
       16 0 35 5
166.66667 0 59 3
      100 1 46 2
      100 1 22 2
      120 1 71 2
133.33333 1 59 3
      200 0 71 1
       60 1 76 1
      120 1 57 2
       60 1 56 2
      120 1 60 2
      130 0 39 2
   243.75 1 40 2
      120 1 53 2
       25 1 62 2
    112.5 1 30 2
       30 1 68 2
      105 1 62 2
      330 1 37 2
      150 1 27 2
      120 1 30 1
      165 0 47 2
       90 0 51 2
      150 1 33 2
end
label values sex sex
label def sex 0 "Female", modify
label def sex 1 "Male", modify
label values educ educ
label def educ 1 "No schooling", modify
label def educ 2 "Primary", modify
label def educ 3 "Secondary", modify
label def educ 4 "Tertiary/University", modify
label def educ 5 "Adult Educ", modify

[Stephen Jenkins]

Look at your trace in detail -- at the top where the tempvars are created, and then where the error arises later on. I have no idea whether the problem is with your dataset or with the code.

The error arises because of issues with a tempvar derived from the dependent (outcome) variable, "yield", not with variable "sex". See below. All values of yield are non-missing, so the tempvar y_s only has 1 category, not two.

. which qregsel
d:\home\stephenj\ado\stbplus\q\qregsel.ado
*! version 1.1 1April2020
*! version 1.2 9September2020 - Simplifies syntax and changes how the estimatio
> n is done
*! version 1.3 11September2020 - Fix some issues regarding collinearity
*! version 1.4 22September2020 - Fix issue with the syntax in the selection eq.

.
.
. ta sex, missing

        sex |      Freq.     Percent        Cum.
------------+-----------------------------------
     Female |         25       25.00       25.00
       Male |         75       75.00      100.00
------------+-----------------------------------
      Total |        100      100.00

. return list    

scalars:
                  r(N) =  100
                  r(r) =  2

.        
. gen y_s = (yield!=.)

. ta y_s, missing

        y_s |      Freq.     Percent        Cum.
------------+-----------------------------------
          1 |        100      100.00      100.00
------------+-----------------------------------
      Total |        100      100.00

. return list

scalars:
                  r(N) =  100
                  r(r) =  1

.
. qregsel yield age educ ///
>         , select(sex) quantile(.5)
__000001 should be binary.
r(198);

end of do-file

r(198);

.

[Co Ar]

Stephen Jenkins am so grateful for the insights. I think the code needs to allow for some space to allow for non missings because not all the time we can have data with missing values. I have contacted the authors to help