*
* Example from section 11.2 on pp 359-361
*
data(org=columns,unit=input) 1 40 state change
 0 -20
 0 -17.5
 0 -13
 0 -12.5
 0 -4.5
 0 -4
 0 -3.5
 0 -2
 0 -1.5
 0 -1
 0 -.5
 0 -.5
 0 0
 0 0
 0 .5
 0 .5
 0 1.5
 0 2
 0 2
 0 2.25
 0 3
 0 4.5
 0 4.5
 0 5.5
 0 6
 0 6.25
 0 8.25
 0 9
 0 10
 0 10.5
 0 12
 0 14.75
 0 34
 1 -7
 1 -6
 1 -2.5
 1 -.5
 1 4
 1 4.5
 1 4.5
*
* The parametric test can be done rather simply by a regression of change on
* constant and the state dummy. The test for equality of means is done by looking
* at the coefficient on the state dummy.
*
linreg change
# constant state
*
* Using RATS, you can do the shuffling of the data set using the BOOT instruction
* with the NOREPLACE option. This draws a SERIES of INTEGERS which, when used as
* subscripts for the original dependent variable, will shuffle the data.
*
compute ndraws=10000
set diffs 1 ndraws = 0.0
do draws=1,ndraws
   boot(noreplace) shuffle
   set ys = change(shuffle(t))
   linreg(noprint) ys
   # constant state
   compute diffs(draws)=%beta(2)
end do draws
*
* The 95% confidence level interval can be obtained by using the %fractiles
* function. We need the .025 and .975 fractiles.
*
compute bounds=%fractiles(diffs,||.025,.975||)
density diffs / dx fx
scatter(style=lines,hgrid=bounds)
# dx fx