[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: time-dependence and serial correlation in event history models
My apologies if anyone gets this twice, but I sent the original more than
24 hours ago and it hasn't shown up yet...
I am running an event history analysis, using logit. The unit of analysis
is the 50 United States and an observation is taken for each state for a
period of 50 months. I am estimating the probability that a state will
experience an event, so a state is censored once the event occurs.
One of my predictors is the cumulative number (or a function of that
number) of states that has previously experienced the event. I am concered
about time dependence and serial correlation (does the effect I find
persist if I control for these?)
I can see three ways of doing this:
1) use robust standard errors for the logit estimation
2) Include a time-counter as one of the predictor variables
3) use the "cluster" option to cluster observations by state.
It appears that "cluster" implies robust SEs, but 2 and 3 are still
I am unclear about which of these to use, and whether it makes any sense to
use more than one of them at the same time.
Any pointers would be greatly appreciated.
Dept. of Political Science
Rhode Island College
600 Mount Pleasant Ave. Providence, RI 02908
Phone: (401) 456-8722
* For searches and help try: