It is common for physiologists to record neuron electrical activity over
time (perhaps under experimental conditions) and perform time series
analysis on these data, regardless of which animal the neuron was
recorded from. I would think that such an analysis mistakes the
experimental unit as the neuron when it should be the animal. Stata
implements panel data time series models, though I have little
experience with them. My feeling is that the panel should be the animal,
and the neuronal recordings should be within-panel observations. For
those economists out there on the list, can you point to literature that
describes the consequences of ignoring the panel, or collapsing over
panel, on parameter estimates and or Type I/II error rates? Literature
that uses Stata would be the best.
