*
* Example 8.3, page 269
*
cal 1955
open data table8-8.prn
data(format=prn,org=columns) 1955:1 1974:1
table
*
set loggdp = log(gdp)
set logemp = log(employment)
set logcap = log(capital)
linreg loggdp
# constant logemp logcap
*
* Testing b2+b3=1.0 (or 1.0xb2+1.0xb3=1.0) using RESTRICT
*
restrict(title="Test of b2+b3=1.0") 1
# 2 3
# 1.0 1.0 1.0
*
* Testing by using SUMMARIZE, which computes the sum of the requested
* coefficients, and the standard error of the sum. The t-stat in the SUMMARIZE
* output is for the null of the sum being zero, while we want to test that it's
* one. We can get the t stat for our test by taking
*
*  (sum-1)/std error
*
* from the output, or by doing the same computation using compute and the %sumlc
* and %varlc variables, which have the sum and variance of the sum.
*
summarize
# logemp logcap
compute tstat=(%sumlc-1)/sqrt(%varlc)
cdf(title="Test of b2+b3=1 using SUMMARIZE output") ttest tstat %ndf
*
* Testing by restricted/unrestricted regressions. We have to use the RSS form,
* since the dependent variable is different.
*
compute rssu=%rss,ndfu=%ndf
set loggdpdemp = log(gdp/employment)
set logcapdemp = log(capital/employment)
linreg loggdpdemp
# constant logcapdemp
compute f=((%rss-rssu)/1)/(rssu/ndfu)
cdf(title="Test of b2+b3=1 using Restricted/Unrestricted Regressions") ftest f 1 ndfu