*
* Example 7.2, page 220
*
cal 1970
open data table7-1.prn
data(format=prn,org=columns) 1970:1 1980:1
*
linreg y
# constant x
*
set logy = log(y)
set logx = log(x)
*
linreg logy
# constant logx
*
* Computing R**2 comparable to the linear regression
*
* Get fitted values for the log model and compute the adjusted residuals:
* y-exp(logfit). The R**2 can be computed by taking 1 minus the ratio of the
* sample variances of the adjusted residuals and the dependent variable. (The
* adjusted residuals, in general, won't have a zero mean - this calculation will
* use the deviations of the adjusted residuals from their mean).
*
prj logfit
set adjresid = y-exp(logfit)
stats(noprint) adjresid
compute resvar=%variance
stats(noprint) y
disp "R**2 Adjusted" 1-resvar/%variance
*
* Analogous calculation, computing an R**2 for the linear regression comparable
* to the one reported for the log-log regression. (The result here differs from
* that cited in the text)
*
linreg y
# constant x
prj levfit
set adjresid = logy-log(levfit)
stats(noprint) adjresid
compute resvar=%variance
stats(noprint) logy
disp "R**2 Adjusted" 1-resvar/%variance