Dear Statalist,
I wanted to confirm if my my derivation and understanding is correct: Suppose we assume that in a dynamic model the idiosyncratic error is not iid, but rather MA(1). In that case, the differenced error would be MA(2). So both the first- and the second-order serial correlation of the *differenced* error term would be non-zero, whereas the third- and higher-order correlations would be zero. Correct so far?
In that case, if my assumption of an MA(1) process in the level of the error is on target, I should expect to see p-values for both the 1st and 2nd order of the -estat abond- test showing non-siginifant test stats, i.e. p-values higher than say 0.10.
My first quetion is: is this right? And secondly, is there any other test through which I can discern if, say the third order correlation is indeed zero as it should be if the MA(1) errors assumption is appropriate? Please note that I am using an estimator with robust standard errors, so the postestimation test -estat sargan- is not available to me.
Many thanks!
Hewan
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
