Good afternoon!
1 I have a bit of a problem with the Survey package, but one that is
more theoretical than practical. If for example on wishes to use some
additional information which comprises of 4 variables with 7, 5,5 and 5
categories (resulting in 875 post-strata if the whole crossing were to
be constructed) and if only the marginal distribution of each variable
is known, would it be better to rake or to calibrate?
2 Is calibrating affected by the cells with 0 frequency, as raking
is?
3 when a sample is already stratified and one wishes to post stratify
by other variables, do the new weights take account of the fact that
the sample is stratified and that sampling weights could be determined;
and what is the algorithm for determining these new weights; are the new
weights determined for each strata taking account of the sampling weights for
each strata ?
Thank you and have a good day!
---------------------------------
[[alternative HTML version deleted]]