regression on filtered data

[ajay pasi]

Is there any disadavntage associated with the filtered regression wherein you you have filtered the dependednt and independ variables at the same time.

e.g., reg Y X1 X2 X3 X4 Gender if Y>10 & (X1>=7 &X1<12) & Gender=="Female" , robust

I have large data set wherein I need to test the relatin of Y and independent variabels in controlled scanario.

Is there any problem with the controls in the formof filters used in stata.?

[Carlo Lazzaro]

Ajay:
no problem, as far as I know:

. sysuse auto.dta
(1978 Automobile Data)

. reg price mpg if price>5000 & (mpg>=17 & mpg<20) & foreign==1

      Source |       SS           df       MS      Number of obs   =         4
-------------+----------------------------------   F(1, 2)         =     18.96
       Model |  25336122.3         1  25336122.3   Prob > F        =    0.0489
    Residual |   2672712.5         2  1336356.25   R-squared       =    0.9046
-------------+----------------------------------   Adj R-squared   =    0.8569
       Total |  28008834.8         3  9336278.25   Root MSE        =      1156

------------------------------------------------------------------------------
       price |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         mpg |    -5033.5   1156.009    -4.35   0.049     -10007.4   -59.59576
       _cons |      96412   20238.41     4.76   0.041     9333.155    183490.8
------------------------------------------------------------------------------

[ajay pasi]

thank yo so much Carlo..)

[ajay pasi]

with the refernce to your response, I would wish to get one more clarification
reg price mpg if price>5000 & (mpg>=17 & mpg<20) & foreign==1, robust what does this (, robust) does? In my own beileif, it takes care of hetrogeiety of variance with some 'weighted' approach.

[Carlo Lazzaro]

Ajas:
under -regress- -robust- option accounts for heteroskedasticity in residual distribution.