Dear Statalisters,
I am trying to use -xtserial- to test for autocorrelation in panel data. It seems that if I run "xtserial y x1-x65", where y is in levels, I have 65 df for the F-test. However, when I run "xtserial ydiff x1-x65", where y = D.y, I have 0 df for the F-test, and thus the F-test cannot be performed. Why is this?
Some further explanation may be in order:
1) Correct me if I am wrong here, but I believe I should run -xtserial- on the differenced (and thus stationary) x's and y's. That is why I want to run the second specification. (Therefore, the variables x1-x65 are also typically differenced versions of the original variables).
2) N = 5876
3) I have 226 clusters
I am pasting snippets of the output from each specification, below, in case it provides any other relevant info that I have not already listed:
. xtserial y x1-x65, output
Linear regression Number of obs = 5876
F( 65, 225) =11377.92
Prob > F = 0.0000
R-squared = 0.8491
Number of clusters (cbsa_fips) = 226 Root MSE = .09969
. xtserial ydiff x1-x65, output
Linear regression Number of obs = 5876
F( 0, 225) = .
Prob > F = .
R-squared = 1.0000
Number of clusters (cbsa_fips) = 226 Root MSE = 0
Many thanks,
Danielle Ferry
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/