forvalues-loop neglects last value, why?

[Mario Schimmels]

Hey guys,
in the context of my master thesis, I need to run some quantile regressions. However, when I run the following section of my code

forvalues qq = .05(.05).95 {
qreg yt_1 nfci yt, quantile(`qq') vce(robust)
}

STATA runs the quantile regressions up to the .9 regression but not the .95 regression. How could this be?

Thanks in advance,
Mario

[Nick Cox]

It is a precision problem. For the explanation and the remedy, see https://www.stata-journal.com/articl...article=pr0051

Let's show more detail.

forval qq = .05(.05).95 { 
di %23.18f `qq' 
} 
   0.050000000000000003
   0.100000000000000006
   0.149999999999999994
   0.200000000000000011
   0.250000000000000000
   0.299999999999999989
   0.349999999999999978
   0.400000000000000022
   0.450000000000000011
   0.499999999999999889
   0.549999999999999933
   0.599999999999999978
   0.650000000000000022
   0.700000000000000067
   0.750000000000000111
   0.800000000000000155
   0.850000000000000200
   0.900000000000000244

Why doesn't Stata (*) show an equivalent of 0.95? Because in effect you are asking for something closer to

. di %23.18f (0.90 + 0.05)
   0.950000000000000067

rather than a direct display. of


.

 di %23.18f 0.95
   0.949999999999999956

You would be better off with

forvalues q = 1/19 {
     local qq = `q' * 0.05 
     qreg yt_1 nfci yt, quantile(`qq') vce(robust)
}

(*) https://www.statalist.org/forums/help#spelling

[Mario Schimmels]

Thanks a lot for this detailed explanation, Nick. Appreciate that much!