[R] Fixed Effects using AER's Tobit function - system is singular

[Jonathaneyer]

I have an unbalanced panel of daily, county data that is naturally bounded at
zero so my intention is to use a tobit. I'm using tobit from the AER
package. There is cyclicality in the data for each pattern that I would like
to control for before I add my variables of interest.

I run the regression:

derp <- tobit(x ~ factor(Month)*factor(County), data = data0, left = 0,
right = Inf)

If I try to summarize the tobit object I receive the following error:

Error in solve.default(L %*% V %*% t(L)) : Lapack routine dgesv: system is
exactly singular: U[1,1] = 0

derp, the tobit object, still has estimated coefficients. The
variance-covariance matrix of derp has some rows and columns that are
completely zero. I think that this may be driving the problem.

I have tried aggregating the data over days in each month and have the same
issue. Similarly, I've aggregated the data over days in each month and
counties in each state and I have the same problem.  I can run the
regression without any problem if I use factor(Month) + factor(County)
instead of the interaction of the effects. 

Any thoughts?



--
View this message in context: http://r.789695.n4.nabble.com/Fixed-Effects-using-AER-s-Tobit-function-system-is-singular-tp4651061.html
Sent from the R help mailing list archive at Nabble.com.