Prob>chi2 equals dot

[Bianca Siswanto]

Hello everyone,

I'm currently running a panel data consists of 124 observations. I want to do LM test and I'm having a problem where my Prob>chi2 = . Here is my regression:
xtreg srd genbod ownbod_w indboc polboc roa ta der age, re

Random-effects GLS regression Number of obs = 124
Group variable: id Number of groups = 59

R-sq: Obs per group:
within = 0.1532 min = 1
between = 0.0037 avg = 2.1
overall = 0.0273 max = 3

Wald chi2(7) = .
corr(u_i, X) = 0 (assumed) Prob > chi2 = .

------------------------------------------------------------------------------
srd | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
genbod | -.0470563 .0899987 -0.52 0.601 -.2234504 .1293379
ownbod_w | -.0403094 2.83212 -0.01 0.989 -5.591162 5.510543
indboc | .2781834 .1231357 2.26 0.024 .036842 .5195249
polboc | -.0151156 .0501099 -0.30 0.763 -.1133292 .083098
roa | -.2119119 .175164 -1.21 0.226 -.5552269 .1314032
ta | 3.85e-10 3.03e-10 1.27 0.203 -2.08e-10 9.79e-10
der | -.0028708 .0119586 -0.24 0.810 -.0263093 .0205677
age | .0007332 .0008396 0.87 0.383 -.0009124 .0023787
_cons | .3315796 .0636695 5.21 0.000 .2067897 .4563695
-------------+----------------------------------------------------------------
sigma_u | .10000396
sigma_e | .06281543
rho | .71707895 (fraction of variance due to u_i)
------------------------------------------------------------------------------

Variable TA (total assets)is in millions (IDR) and the rest of the variables are relatively small. My data looks something like this:
id indboc polboc roa ta der age ownbod_w
1. .25 0 .05 2.7e+07 .38 30 0
2. .25 0 .01 2.7e+07 .42 31 0
3. .33 .67 .02 3.0e+07 3.79 58 1.00e-05
4. .33 1 .02 3.7e+07 4.34. 59 1.00e-05



I have read several answers on Statalist regarding to this matter, but I can't find one that matches my exact problem. Is there any way I can still use TA without having to do log/ln? Because if I generate lnTA/logTA, I will have multicollinearity problem. I'm using Stata 14 for Mac. Thank you very much in advance.

Regards,
Bianca

[Nick Cox]

I would divide total assets by a million or perhaps more just to avoid very small coefficients. Whether using its logarithm would be better is hard to say but whence the multicollinearity? I doubt that anyone would advocate using both total assets and its logarithm as predictors. It's hard to read results not presented as CODE (please see FAQ Advice) but total assets doesn't seem to be helping much. Quite why may well depend on what else is in the model.

[Bianca Siswanto]

Got it. I have divided total assets by a million and it worked. Thank you very much for your help