Typo in xtile.ado with option altdef

[Mauricio Caceres]

Line 135 of xtile.ado appears to contain a typo. The line reads

    args wt x newvar nq altdev

However, as best I can tell it should read

    args wt x newvar nq altdef

To see the incorrect behavior, consider

set rmsg on
clear
set obs 10
gen x = _n
xtile x1 = x, nq(7)
xtile x2 = x, nq(7) altdef
pctile p1 = x, nq(7)
pctile p2 = x, nq(7) altdef
xtile xp1 = x, c(p1)
xtile xp2 = x, c(p2) altdef
l

This shows

     +------------------------------------------+
     |  x   x1   x2   p1         p2   xp1   xp2 |
     |------------------------------------------|
  1. |  1    1    1    2   1.571429     1     1 |
  2. |  2    1    1    3   3.142857     1     2 |
  3. |  3    2    2    5   4.714286     2     2 |
  4. |  4    3    3    6   6.285714     3     3 |
  5. |  5    3    3    8   7.857143     3     4 |
     |------------------------------------------|
  6. |  6    4    4    9   9.428572     4     4 |
  7. |  7    5    5    .          .     5     5 |
  8. |  8    5    5    .          .     5     6 |
  9. |  9    6    6    .          .     6     6 |
 10. | 10    7    7    .          .     7     7 |
     +------------------------------------------+

x1 == x2, but this should not be the case. Using cutpoints and pctile we get the different (and expected) results. I only have access to Stata for Unix, but this is present in 13, 14, and 15.

[Daniel Bela]

I can confirm this is also true in xtile.ado under Stata 15 for Windows -- as you would expect, as the system ado files are platform-independent, I guess. I think Mauricio Caceres spotted a bug here.

Does someone from StataCorp read here and can confirm this? If not, you should probably raise a support ticket with Stata technical support.

Regards
Bela

[Nick Cox]

I've sent the reference of this thread to StataCorp technical support.

[Alan Riley (StataCorp)]

I can confirm that this is indeed a typo in the code. We will fix it in a future update.