Hi everybody,

I wondered whether it does make sense to condition on the lagged dependent variable when preprocessing a dataset.

Ho et al. (2007) write:
"To ensure that selection during preprocessing depends only on Xi (to prevent inducing bias), the outcome variable Yi should not be examined during the preprocessing stage. As long as Yi is not consulted and is not part of the rule by which one drops observations, preprocessing cannot result in stacking the deck one way or another."
So does this mean the lagged DV should not be included?

Best,
Tobias