Dear all,
I am writing a code on some flexible estimation method and I wanted to use "qreg" as preliminary estimator. However I noticed that coefficients associated to factor variables are not (always) the same wether you use or not the xi: prefix (even if cathegories are the same).

See for example rep78 in the following example:

. sysuse auto
(1978 Automobile Data)


. qreg price mpg i.rep78 headroom weight
Iteration 1: WLS sum of weighted deviations = 57526.806

Iteration 1: sum of abs. weighted deviations = 87541.5
Iteration 2: sum of abs. weighted deviations = 58455.934
Iteration 3: sum of abs. weighted deviations = 53565.321
Iteration 4: sum of abs. weighted deviations = 53450.253
Iteration 5: sum of abs. weighted deviations = 53117.82
Iteration 6: sum of abs. weighted deviations = 52995.225
Iteration 7: sum of abs. weighted deviations = 52959.223
Iteration 8: sum of abs. weighted deviations = 52921.022
Iteration 9: sum of abs. weighted deviations = 52889.25

Median regression Number of obs = 69
Raw sum of deviations 65163 (about 5079)
Min sum of deviations 52889.25 Pseudo R2 = 0.1884

------------------------------------------------------------------------------
price | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
mpg | -70.95877 112.518 -0.63 0.531 -295.9524 154.0348
|
rep78 |
2 | 27.90316 2596.398 0.01 0.991 -5163.916 5219.723
3 | 61.36817 2386.773 0.03 0.980 -4711.28 4834.016
4 | 1584.145 2434.712 0.65 0.518 -3284.363 6452.653
5 | 2078.078 2523.327 0.82 0.413 -2967.627 7123.782
|
headroom | -154.3778 532.8143 -0.29 0.773 -1219.806 911.0503
weight | 1.272579 .8580482 1.48 0.143 -.4431944 2.988353
_cons | 2732.625 5098.427 0.54 0.594 -7462.31 12927.56
------------------------------------------------------------------------------

. xi: qreg price mpg i.rep78 headroom weight
i.rep78 _Irep78_1-5 (naturally coded; _Irep78_1 omitted)
Iteration 1: WLS sum of weighted deviations = 57526.806

Iteration 1: sum of abs. weighted deviations = 87541.5
Iteration 2: sum of abs. weighted deviations = 58455.934
Iteration 3: sum of abs. weighted deviations = 53565.321
Iteration 4: sum of abs. weighted deviations = 53450.253
Iteration 5: sum of abs. weighted deviations = 53117.82
Iteration 6: sum of abs. weighted deviations = 52995.225
Iteration 7: sum of abs. weighted deviations = 52959.223
Iteration 8: sum of abs. weighted deviations = 52921.022
Iteration 9: sum of abs. weighted deviations = 52889.25

Median regression Number of obs = 69
Raw sum of deviations 65163 (about 5079)
Min sum of deviations 52889.25 Pseudo R2 = 0.1884

------------------------------------------------------------------------------
price | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
mpg | -70.95877 112.518 -0.63 0.531 -295.9524 154.0348
_Irep78_2 | 733.5532 2596.398 0.28 0.778 -4458.266 5925.373
_Irep78_3 | 767.0182 2386.773 0.32 0.749 -4005.63 5539.666
_Irep78_4 | 2289.795 2434.712 0.94 0.351 -2578.713 7158.303
_Irep78_5 | 2783.728 2523.327 1.10 0.274 -2261.977 7829.432
headroom | -154.3778 532.8143 -0.29 0.773 -1219.806 911.0503
weight | 1.272579 .8580482 1.48 0.143 -.4431944 2.988353
_cons | 2026.975 5098.427 0.40 0.692 -8167.96 12221.91
------------------------------------------------------------------------------



I immagine I am doing a silly error somewhere but I cannot flag it. This does not happen with rreg for example.

Predicted y_hats are not identical either.

Thank you for the help

Best

Vincenzo