st: How to test coefficients obtained from Fama-MacBeth regressions across groups?

[LI Mengjia <limengjia626@163.com>]

Hi statalists,

I want to test all the coefficients obtained from Fama-MacBeth regressions across 2 groups.

I checked up with the FAQ files (http://www.ats.ucla.edu/stat/stata/faq/compreg2.htm and http://www.ats.ucla.edu/stat/stata/faq/compreg3.htm)and some similar questions raised before(http://www.stata.com/statalist/archive/2011-11/msg00398.html), but find the way they offered (most say adding a dummy and its interaction with other Xs) unable to solve my problem.

I just want to test whether the coefficients are significantly different when x3>0 and x3<=0, where x3 remains constant at any given t. So I created a dummy, group, group=1when x3>0 and group=0 when x3<=0. Thus, the group dummy at any given time t takes the same value (e.g. for t=1,2, group=1, for t=3,4, group=0). Therefore, when -xtfmb- runs the first step (cross-section regression at every t), the group dummy has no effect and will be simply going to the constant term. 

I also tried -suest-, but just like the above mentioned archive file, it returns "was estimated with a nonstandard vie (Fama-Macbeth)".

Can anyone kindly tell me how to deal with this problem? My data structure looks like:

Code	time		y	x1	x2	x3	group
001		1		15	8	2	0.78		1
001		2		11	5	3	0.65		1
002		1		23	14	6	0.78		1
002		2		31	18	10	0.65		1
002		3		19	7	11	-0.23	0
003		1		78	43	20	0.78		1
003		2		69	45	10	0.65		1
003		3		55	32	26	-0.23	0
003		4		83	54	41	-0.73	0

Best regards and many thanks,
Mengjia

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/