[R-sig-ME] weights-option in lmer
Ben Pelzer
b.pelzer at maw.ru.nl
Wed May 26 21:26:53 CEST 2010
Dear all,
I have to related questions concerning the weights-option in lmer.
1) Imagine a two level design, like respondents in a number of
countries. From each country one has a sample and the percentages of
males and females in each country differ from the population percentages
for that country. Can one use the weights-option in lmer to obtain a
weighted sample for which the (weighted) male/female sample-percentages
are equal to the population percentages, to achieve more representative
results?
2) If one has aggregated data-records, each record containing the values
of the dependent and independent variables and also the number of
lower-level cases in a variable named 'nofcases', can the '
weight=nofcases' option in lmer then be used to 'disaggregate' the
data? And are the resulting standard errors then correct, i.e. related
to the total number of lower-level (disaggregated) cases?
If the answer to these question is 'yes', which I hope, does that also
hold for the nonlinear, especially the binomial family models?
Thanks for any help!!
Ben.
More information about the R-sig-mixed-models
mailing list