* Use CEOSAL2.DTA
regress lsalary lsales lmktval ceoten
* residual MSE = 0.254791732
predict logyhat
generate ytilda = exp(0.254791732/2)*exp(logyhat)
generate ssr_tilda = 177*(salary - ytilda)^2
summarize ssr_tilda
* ssr_tilda = 4.62e+07
regress salary sales mktval ceoten
* sst = 60765964.7
* R2_yhat = 0.2013  
scalar R2_tilda = 1 - 46200000/60765964.7
display R2_tilda
* R2_tilda = 0.23970597
* Therefore, the "lsalary lsales lmktval ceoten" equation is preferred
* to the "salary sales mktval ceoten" equation.