*
* Example BMS.DAT from pp 475-479
*
open data bms.dat
data(format=free,org=columns) 1 100 sales advertising comp pop
*
* First stage regressions
*
linreg sales
# constant comp pop
linreg advertising
# constant comp pop
*
* Two Stage Least Squares. The first stage regressions shown above aren't
* actually used, because the full calculation is done with a single LINREG
* instruction with the INSTRUMENTS options. This is preceded by an INSTRUMENTS
* instruction which lists the predetermined variables.
*
instruments constant comp pop
linreg sales ;* OLS estimates
# constant advertising comp
linreg(instruments) sales
# constant advertising comp
*
linreg advertising
# constant sales pop
linreg(instruments) advertising
# constant sales pop
*
* Hausman test done using auxiliary regressions.
*
linreg advertising
# constant pop comp
set v1 = %resids
linreg sales
# constant advertising comp v1
exclude(title="Hausman Specification Test for endogeneity of advertising")
# v1
*
* Hausman test doing using the RegWuTest procedure. (The Wu test is a special case
* of the Hausman test applied to 2SLS).
*
linreg(instruments) sales
# constant advertising comp
@RegWuTest