Answers are grounded in this paper and cite their sources.
The authors describe methodological issues of panel analysis designs; in particular, autocorrelation of errors across waves. They recommend constructing a “pooled” model of data from all waves to deal with design and estimation problems, and present results from a simulation of the behavior of alternate estimators for pooled models.
No takes yet. Share an insight, caveat, or question.
Hannan et al. (1977) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: