Dear all;
I'm looking for some advice regarding the following idea:
Let's say that I have a sample of y-values randomly taken from a
population and I want to compute the mean of y and its confidence
intervals but without assuming any particular distribution (I'm
assuming that the mean of this sample is a good indicator of the mean
of the population of y's). As far as I know we can use a nonparametric
bootstrap analysis approach to do something like this.
Now, let's say that instead of having to measure "y", I can
measure
"x" because is easier. Moreover I have a model that relates y and x,
so I can predict the "y" giving the set of observed x. At the end of
the day I will have yhat=(y1-hat,...,yn-hat)' which is the vector of
predicted y-values.
Here the is question: Does it make any sense to try to calculate the
mean of the predicted "y's" and its CI by using a bootstrap
analysis?
Am I violating any assumptions for that kind of analysis? (maybe the
independence of the samples?)
Sorry if this is a dumb question but I would like to have a different opinion
Thanks in advance
PM
