Hi
I used Hausman test to choose between a fixed effects and random effects but i am very surprised by results:
First , the value if the chi2 is too big and its sign depends on the order of commands user in stata
1) if i started with random commands in stata are:
xtreg y x, random
estimates store random
xtreg y x, fixed
hausman random
---- Coefficients ----
| (b) (B) (b-B) sqrt(diag(V_b-V_B))
| random . Difference S.E.
-------------+----------------------------------------------------------------
rating | .0142675 .0096562 .0046113 .
varautor | -.0005397 -.0002915 -.0002482 .0001096
def | .1071089 .0880998 .0190091 .
csr | .037553 .0331016 .0044514 .
anc | -.0007706 -.0007352 -.0000354 .
tx | .0038553 .0060001 -.0021448 .
spread | .0083779 -.0217386 .0301165 .
ech | .0004364 .000571 -.0001346 .
varpibg | .0260169 .0341001 -.0080832 .0074378
f1 | .1289178 .0356372 .0932806 .
f2 | .0457407 .0389245 .0068162 .0002615
f3 | .0374901 .033304 .0041861 .0003952
f4 | -.017298 -.0144426 -.0028554 .
---------------------------------------------------------------------------
b = consistent under Ho and Ha; obtained from xtreg
B = inconsistent under Ha, efficient under Ho; obtained from xtreg
Test: Ho: difference in coefficients not systematic
chi2(13) = (b-B)'[(V_b-V_B)^(-1)](b-B)
= -1383.63 chi2<0 ==> model fitted on these
data fails to meet the asymptotic
assumptions of the Hausman test;
see suest for a generalized test
2-If i start with fixed commands in stata are :
xtreg y x, fixed
estimates store fixed
xtreg y x, random
hausman fixed
---- Coefficients ----
| (b) (B) (b-B) sqrt(diag(V_b-V_B))
| fixed . Difference S.E.
-------------+----------------------------------------------------------------
rating | .0096562 .0142675 -.0046113 .000271
varautor | -.0002915 -.0005397 .0002482 .
def | .0880998 .1071089 -.0190091 .0069914
csr | .0331016 .037553 -.0044514 .0003885
anc | -.0007352 -.0007706 .0000354 .0000515
tx | .0060001 .0038553 .0021448 .0002077
spread | -.0217386 .0083779 -.0301165 .0014616
ech | .000571 .0004364 .0001346 .0000186
varpibg | .0341001 .0260169 .0080832 .
f1 | .0356372 .1289178 -.0932806 .0042425
f2 | .0389245 .0457407 -.0068162 .
f3 | .033304 .0374901 -.0041861 .
f4 | -.0144426 -.017298 .0028554 .0010139
----------------------------------------------------------------------------
b = consistent under Ho and Ha; obtained from xtreg
B = inconsistent under Ha, efficient under Ho; obtained from xtreg
Test: Ho: difference in coefficients not systematic
chi2(13) = (b-B)'[(V_b-V_B)^(-1)](b-B)
= 1383.63
Prob>chi2 = 0.0000
(V_b-V_B is not positive definite)
Second i got the remark V_b-V_B is not positive definite , i don't know what does it mean neither how to deal with this problem
third , when comparing the fixed estimates with random estimates i didn't see a big differences in value except for the variable spread which sign has changed .
I would be very greatful if someone help me in resolving those problems. I ve alredy tried the commnd suest but xtreg didn't accept to undertake the option score() which is necessary.
Best whises
Hanaa
*
* 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/
