I'm currently working with a longitudinal dataset, the Kauffman Firm Survey.
The survey tracks about 5000 firms starting from 2004 - 2009. Firms die out over the years.
It has both cross-sectional weights and longitudinal weights.
I've checked out Lumley's Complex Surveys using R, but the section on repeated samples does not provide guidance on the use of different weights. I also have Levy and Lemshow's Sampling of Populations but its section on weighting says longitudinal weights is outside the scope of the book. The most explicit advice I've found on the matter is here: http://help.pop.psu.edu/help-by-statistical-method/weighting/Introduction%20to%20survey%20weights%20pri%20version.ppt , slide 23.
From this power point, I would assume the following use cases:
a) Let’s say I want to get the mean revenue from years 1,2,3, but only for firms that survived through year 3, then I would use the longitudinal weight.
b) However, if I want to get the mean revenue for all firms from years 1,2,3 (including those that died out), I would use the cross section baseline weight for year 1 – NOT the longitudinal weights for that year, which is what I would have presumed needed to be used.
c) Lastly, if I only wanted to look at the mean revenue of firms from year 3 (not including years 1 and 2), I would use cross-section weight for year 3.
Questions:
1. Are these statements correct?
2. What are some good resources for the correct use of longitudinal (panel) weights?
