re:RE: st: FW: comparing xtregar coefficients across models

[Christopher Baum <kit.baum@bc.edu>]

<>
Sorry, forgot to mention that I also reversed the order of running the two models i.e.-

Run model 2.
Save the beta in a scalar as:
. scalar b_pastyearseaavgwq=_b[pastyearseaavgwq]
Run model 1.
Run test as:
. test _b[lagseaavgwqfoia04avg3]=b_pastyearseaavgwq

The test statistic shows that the two coefficients are not significantly different from each other, in both cases. So, can I conclude then that the coefficients are not statistically different?


NO. You cannot test whether two random variables are equal to one another by setting one of them to a fixed value, no matter how you do it.

suest is quite restrictive in terms of what models can be employed (e.g., xtreg does not work either). If you can fit the models with explicit dummies (using i.panelvar regressors), estimate each model with -regress- and then use suest. But the logic of comparing an estimated coefficient to a fixed value does not correspond to testing whether these two estimated coefficients can be distinguished from each other by the data.

Kit Baum   |   Boston College Economics & DIW Berlin   |   http://ideas.repec.org/e/pba1.html
                              An Introduction to Stata Programming  |   http://www.stata-press.com/books/isp.html
   An Introduction to Modern Econometrics Using Stata  |   http://www.stata-press.com/books/imeus.html




*
*   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/