I've been doing some work looking at how to best estimate what a local level population looks like.
The way that I've eventually come round to thinking about it is to mash together individual level data that is sparsely spread around the the total area e.g. the UK, and aggregated data which covers the whole area but doesn't go down to the specifics of one person.
This is an approach that was used some time before I thought of it by several authors, notably C Jackson whose work I've been reading up on and what he call hierarchical related regression. He's also written a handy R package that I plan on looking in to very soon.
Separately, I've been looking at some work on capture-recapture methods to measure hard-to-reach populations. It seems to me that this is a way of combining several individual level data sets into one, with an extra group of "missed" individuals.
I also saw something on the Network Scale-up method that's an approach to reach hard-to-find populations by first estimating the networks of randomly selected people and then estimating the sizes of the hidden population. I don't really know anything about this method so will be reading up on it soon.
Perhaps the capture-recapture and the HRR methods could work together some how, formally or informally?
No comments:
Post a Comment