[Naeem BhaBha]

Dear Wouter Wakker thank you very much for the prompt response. Please can you explain with the example. I tried my best but post 2 and 3 are not working in my case.
reg wroa CSR LOG_AT_US wsale_gr wche_at wbook_td i.fyear i.ff48i if CEOage>50.5, cluster(GVKEY1)
reg wroa CSR LOG_AT_US wsale_gr wche_at wbook_td i.fyear i.ff48i if CEOage<=50.5, cluster(GVKEY1)

From the above codes, I got the following output

	Age > 50.5	Age <= 50.5
CSR	0.013***	0.014***
	(3.69)	(4.91)
SIZE	-0.001	-0.001
	(-0.31)	(-0.63)
SALES_GR	0.071***	0.076***
	(8.94)	(9.26)
CASH/TA	-0.032	-0.023
	(-1.56)	(-1.10)
LEV	-0.099***	-0.096***
	(-5.45)	(-5.41)
Constant	0.241***	0.229***
	(7.81)	(12.66)
Industry Fixed Effects	Yes	Yes
Year Fixed Effects	Yes	Yes
N	6,102	6,116
Adj R2	0.178	0.186

My question is which test I should perform, which tells me that CSR coefficients are significantly different from each other.

[Wouter Wakker]

Method from #2:

gen highage = CEOage > 50.5
reg wroa (c.CSR c.LOG_AT_US c.wsale_gr c.wche_at c.wbook_td i.fyear i.ff48i)##i.highage, cluster(GVKEY1)

Method from #3:

reg wroa CSR LOG_AT_US wsale_gr wche_at wbook_td i.fyear i.ff48i if CEOage>50.5
estimates store M1
reg wroa CSR LOG_AT_US wsale_gr wche_at wbook_td i.fyear i.ff48i if CEOage<=50.5
estimates store M2
suest M1 M2, cluster(GVKEY1)
lincom [M1_mean]CSR - [M2_mean]CSR

[Naeem BhaBha]

Sir Wouter Wakker Thank you very much. This is the excellent way you helped me but I stuck with such error, please what I need to do to

suest M1 M2, cluster(GVKEY1)
M1 was estimated with cluster(GVKEY1).
re-estimate without the cluster() option, and
specify the cluster() option with suest.
r(322);

[Wouter Wakker]

Error messages in Stata can be very uninformative sometimes, but I think in this case it tells you quite a lot. It is telling you to estimate the models themselves without the cluster option, and do the standard error adjustment later when you call the suest command,

Also, from help suest:

2. Estimation should take place without the vce(robust) or vce(cluster clustvar) options. suest always computes the robust estimator of the (co)variance, and suest has a
vce(cluster clustvar) option.

[Naeem BhaBha]

Thank you very much sir Wouter Wakker for your wonderful help.

[Juan Martin Rinaldi]

At the risk of annoying people for self-promotion: xlincom from SSC may be of help here. test does joint significance tests when you give it multiple expressions, but xlincom can get coefficients, standard errors etc. for multiple linear combinations. You can still do a joint significance test afterwards.

. sysuse auto
(1978 Automobile Data)

. qui regress price mpg headroom if foreign, absorb(rep)

. est sto foreign

. qui regress price mpg headroom if !foreign, absorb(rep)

. qui suest foreign .

.
. xlincom (foreign_mean:mpg - foreign_mean:headroom) (foreign_mean:mpg - _LAST_mean:mpg), post

lc_1: foreign_mean:mpg - foreign_mean:headroom = 0
lc_2: foreign_mean:mpg - _LAST_mean:mpg = 0

------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lc_1 | -143.9501 575.6055 -0.25 0.803 -1272.116 984.2161
lc_2 | 262.1829 175.1901 1.50 0.135 -81.18333 605.5491
------------------------------------------------------------------------------

. test (lc_1) (lc_2)

( 1) lc_1 = 0
( 2) lc_2 = 0

chi2( 2) = 2.38
Prob > chi2 = 0.3048

This is really useful! Thank you very much for sharing. I have a theoretical doubt. The larger the amount of joint hypothesis that you want to test, the larger the degrees of freedom that the chi2 is going to have, so the more penalizing the test will be (meaning that it will be more likely to reject the null hypothesis of all coefficients estimators being jointly different from zero). I've tested this on a Probit and obtain statistically significant individual differences of estimators between two models in 3/12 parameters, but all together we reject them being equal at less than 1% level of significance. Is there any way to treat this? What is this signaling, exactly?

[Dan Mosman]

Well, having read previous posts on -suest-, you are presumably aware that you can't do this because -suest- does not support -xtreg-.

However, you can get the comparison you are looking for as follows. I will assume that IV1, IV2 and CV1 are continuous variables, and that CV2 and CV3 are discrete, to illustrate the approach. If that is not the case, you will need to modify the code accordingly.

xtreg DV i.S##(c.(IV1 IV2 CV1) i.(CV2 CV3 )), fe vce(cluster panelvar)
lincom 1.S#IV1 + 1S.#IV2

The -xtreg- command above uses an interaction between S and all of the predictors of the model, thus completely emulating two separate subset regressions. Since the 1.S#whatever terms represent differences between the S = 0 and S = 1 coefficients, the -lincom- command calculates the difference between the S = 0 and S = 1 values of _b[IV1] +_b[IV2].

If you are not familiar with the i. and c. prefixes and the ## operator, read -help fvvarlist- so you will learn about one of Stata's very best features!

Clyde Schechter , could you expound on this if say IV2 is collinear with the panelvar? I would like to run a similar test but use fixed effects in my individual regressions, but if I don't use fixed effects in the test regression to compare differences then I don't feel like I'd be comparing the right thing.

[Clyde Schechter]

I don't see a problem if IV2 is colinear with the panel variable. The regression output will omit the IV2 term, but the interaction term with S will still be there (unless there is some other colinearity involving S), and the result you need depends only on the interaction coefficients.

[Dan Mosman]

Clyde Schechter I apologize I misspoke in my previous post. The issue is with S. So my S variable has a colinearity issue with the panel variable. I've tried the other workaround offered in this thread by Joro Kolev but I have a lot of panel observations which would lead me to use areg and employ his method. However, suest doesn't support areg. Any thoughts on this?

[Clyde Schechter]

OK, we all misspeak at time, so no worries.

I was concerned about the possibility that both S and IV2 were colinear with the panel effect. But if it's just S and not IV2, again, I don't see a problem here. Just proceed with the interaction approach in -xtreg, fe-. It shouldn't be a problem, and this is a standard approach to cross-level interactions. Just make sure that none of the other variables in the model are also invariant within panel. If that's the case, you are toast.