*
* Example of serial correlation on pp 103-106
*
open data icecream.dat
calendar(daysperentry=28) 1951:3:18
data(format=prn,org=columns) 1 30 cons income price temp time
set temp100 = temp/100
graph(footer="Figure 4.2 Ice Cream Consumption, Price and Temperature",key=below) 3
# cons
# price
# temp100
*
linreg cons / resids
# constant price income temp
prj fitted
graph(footer="Figure 4.3 Actual and Fitted Values for Ice Cream Consumption") 2
# fitted
# cons
*
* Breush-Godfrey test. If there's a question about the exogeneity of the
* regressors, you would include them in this auxiliary regression.
*
linreg resids
# resids{1}
cdf(title="Breusch-Godfrey Test-Exogenous Regressors") chisqr %trsquared 1
*
* Estimation with AR1 errors. The default method for choosing the "rho" is the
* Hildreth-Lu grid search procedure. You can use the METHOD=CORC option on AR1 if
* you want to use Cochrane-Orcutt. It will give almost identical results in this
* case.
*
ar1 cons
# constant price income temp
*
* Rerunning the OLS regression with current and lagged temp. You show lags with
* the notation {list of lags}. Note that lags are shown as **positive** values in
* the list - not negative.
*
linreg cons
# constant price income temp{0 1}