Hi All
Does anyone have any references on computing the variance of the differences
between two overlapping samples for binomial proportions, specifically when
one sample (say n_y) is a subset of a much larger sample (n_x).
Kish notes in the book "Survey Sampling" that :
Var(px - py) = (px -px^2)/n_x +(py -py^2)/n_y - 2*(p(xy) -p(x)p(y))/(n_x)
where the proportions with the attributes are px among the n_x and py among
the n_y, and p(xy) denotes the proportion having the attribute in both
samples.
I am not sure how to compute p(xy)? Any suggestions - or links to a simple
example would be appreciated.
regards,
Vince
|