Hi,
I have one dataset holding information on variable X and another unrelated
dataset holding information on variable Y. The variables are independent. I
want to calculate 95% confidence intervals for the mean of X multiplied by
the mean of Y.
One thought is to run separate mean bootstaps on each dataset, save each
bootstrap repetition and then multiply these saved repetitions together,
then order them and get 95% intervals in the usual fashion. If each
bootstrap is 100 runs, do I have to form all 10,000 X. Y permutations or can
I get away with only 100 calculations - assigning the first Y run to the
first X run, 2nd Y run to 2nd X run, etc?
I might be making things too hard for myself - but I can't see an easy way
of analytically computing the confidence intervals by invoking normaility
and independence?
I could apply monte carlo techniques and take draws from the two normal mean
distributions and simply multiply each draw together and then rank them to
give the confidence intervals. This actually looks the easiest way.
Any advice is much appreciated, especially grateful if anyone has an example
of any Stata code for such bootstrapping or monte carlo's.
Many thanks,
Stephen
Stephen Kay
Adelphi Group Products
|