Creating subset of population sample and removing observations outside of the sample

[Lena Gan]

Hi
I ran a descriptive statistics command on my subset population sample and the number of observations tallied up. However, when I ran regression using my predictors, the number of observations did not tally up. It was uneven, including the observations in the outcome. Are there commands I should use to tell Stata to drop all the observations that are outside my population sample so that the analysis and tests I run only pertain to the population sample? Thank you.

[Carlo Lazzaro]

Lena:
if one or more of your predictors have missing values, the observations will be listwise deleted by Stata.
If this is not your case, an -if- qualifier (instead of -drop-ping data that can be useful again) is the simplest way to go:

. sysuse auto.dta
(1978 Automobile Data)

. sum displacement

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
displacement |         74    197.2973    91.83722         79        425

. g flag=1 if displacement>=r(mean)
(38 missing values generated)

. replace flag=0 if flag==.
(38 real changes made)


. sum price if flag==1

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
       price |         36    6836.361    3434.593       3299      15906

. regress displacement i.foreign if flag==1
note: 0.foreign omitted because of collinearity

      Source |       SS           df       MS      Number of obs   =        36
-------------+----------------------------------   F(0, 35)        =      0.00
       Model |           0         0           .   Prob > F        =         .
    Residual |      132716        35  3791.88571   R-squared       =    0.0000
-------------+----------------------------------   Adj R-squared   =    0.0000
       Total |      132716        35  3791.88571   Root MSE        =    61.578

------------------------------------------------------------------------------
displacement |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
     foreign |
   Domestic  |          0  (omitted)
       _cons |   277.3333   10.26305    27.02   0.000     256.4982    298.1684
------------------------------------------------------------------------------

.

Eventually:

summarize if e(sample)

after regression can help you in sniffing out the cuplrit(s) (if any).