## Friday, May 25, 2007

### Pi sub i's as a probability vector for sampling w/o replacement.

I am not sure what Dr. Fellegi meant in his Ph.D. by saying the probabilities of a sample without replacement are to be represented by $\pi_{i}$ vector. We have the equal probabilities of $\frac{1}{N}$ and I would think probabilities of $\frac{1}{(N-n)}$ as sampling without replacement. I thought this was trivial not something one could write a Ph.D. about. Ok I am still assuming equal probabilities and certainly things are more interesting with unequal probabilities.

I do see his experiences with Statistics Canada and thus awareness of respondent burden cropping up early in his dissertation.