At 01:43 AM 11/10/2007, Jason Burke wrote:
>I am interested in a sampling WOR method, whereby the sample drawn
>will match user-specified priors.
I was hoping somebody would know this better than I do, and could
respond with exactly what you want.
See if I have this right: There is a set of N object o(i),i=1,n. With
each is associated a *prior probability* p(i), 0<=p(i)<=1.
Do you also have that
sum(i=1,N)(p(i))=1
Then (I'm still surmising) you want a sample S containing K of the N
objects, where you specify K; where
P(o(i) is in S) = p(i)*K/N
I think this can be done by a variation on either random sorting or
the 'k/n' method (I'm more confident of the latter), but I'm not
working out either before checking that this is what you want.
-Onward,
Richard
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
