I have a panel data set with a non-stationary regressand and a
non-stationary regressor. I've used the plm package to do a fixed effects
regression of a long-run (i.e. cointegrating) relationship between these
non-stationary variables and I've used plm's purtest to do an Engle-Granger
test that the residuals of that regression are indeed stationary. Now, I
need to run an error correction model (ECM) to identify the dynamics -- the
speed at which the regressand converges to the level implied by that
cointegrating relationship in response to a shock to the regressor. The plm
package does not seem to be capable of estimating an ECM (unless I missed
something). Packages like urca seem to offer ECMs but only for pure time
series (i.e. they do not handle panel data). I've searched but have not
found other options in r. Any suggestions out there? I've come up with 3
options:
1. Try to code the maximization of the appropriate ECM's likelihood
2. Try to transform an ECM into the sort of dynamic gmm model that can be
handled by plm (or even by a package not designed for econometrics like
nmle, lme4, or lmer)
3. Try to trick urca (or another package capable of estimating an ECM) into
respecting the panel structure of the data by inserting na observations
into the "time series" every time the panel rolls over to the next
individual?
Thanks for any help,
Bentley
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
