In order to correct the autocorrelation problem, then transform the
lnx and lny into
lnx* and lny* by using the Durbin-Watson statistics to estimate p.
First use the following command line to generate lnx* and lny*
"genr lnxstar = lnx - (0.400089 * lnx(-1))"
"genr lnystar = lny - (0.400089 * lny(-1))"
And then regress lnystar on lnxstar, here are the results:
