Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: RE: huge test statistic with -logit-


From   "Nick Cox" <[email protected]>
To   <[email protected]>
Subject   st: RE: huge test statistic with -logit-
Date   Mon, 7 Dec 2009 15:24:45 -0000

When a model gets to be this fragile, the very general possibilities are
you are trying to include more predictors and/or use a model that the
data won't bear. Either way, retreat to something much simpler and then
build up to see where the problems enter. 

Nick 
[email protected] 

Sue

I'm seeing some huge test statistics that I can't seem to explain. I
run -logit- on a set of regressors (some of them dummies, most of them
not; some country level, some firm level variables), and I also
interact all the regressors with another dummy and include those as
well. I then test for the significance of all the interaction terms.
The test statistics under this baseline model are in a reasonable
range; however, when I add just one more country level variable (which
is continuous), the test statistic on the interaction term goes crazy
and hits a number between 4-7 figures. Does anyone have an idea why
this might be happening? I'm showing part of the output below to give
you an idea:


africa (d)	-177.1579***
	(2.6441)
small (d)	-0.6886***
	(0.0987)
_IafrXsmall_1 (d)	-0.3126
	(0.2492)
medium (d)	-0.1960***
	(0.0574)
_IafrXmediu_1 (d)	-0.1700
	(0.1455)
foreign (d)	-0.2349***
	(0.0789)
_IafrXforei_1 (d)	0.2128
	(0.2149)
exporter (d)	0.2909***
	(0.0487)
_IafrXexpor_1 (d)	-0.4631***
	(0.1264)
manufacturing (d)	0.5076***
	(0.1562)
_IafrXmanuf_1 (d)	-0.9735***
	(0.2952)
services (d)	-0.2403
	(0.1511)
_IafrXservi_1 (d)	-0.9857*
	(0.5771)
gdp_gr	6.8111
	(5.1657)
_IafrXgdp_g_1	4.0075
	(6.2356)
inf	-1.6883
	(1.5576)
_IafrXinf_1	49.0775***
	(1.6506)
kkm	0.8556**
	(0.3435)
_IafrXkkm_1	-13.0895***
	(0.3669)
ca_gdp	1.2100
	(2.7650)
_IafrXca_gd_1	128.8948***
	(3.3821)
bc	-1.3390
	(0.8460)
_IafrXbc_1	75.3131***
	(1.8706)
fos	-2.1180***
	(0.7812)
_IafrXfos_1	96.0801***
	(1.3381)
gdp_pc	-0.0085
	(0.2651)
_IafrXgdp_p_1	-6.8328***
	(0.2706)
ln_pd	0.2728*
	(0.1458)
_IafrXln_pd_1	7.3811***
	(0.2421)
natres	-0.2812
	(0.3640)
_IafrXnatre_1	133.4017***
	(7.8873)
ln_population	-0.0640
	(0.1086)
_IafrXln_po_1	24.6676***
	(0.2401)
ofc (d)	-0.9053
	(0.9095)
pop_gdp	-0.0151
	(0.2987)
_IafrXpop_g_1	-705.4406
	(0.0000)
sec2prim_enrol	0.4339
	(0.5475)
_IafrXsec2p_1	133.0970***
	(1.3934)
new_geobrpen	-0.1639                  --> The variable that was added
to the baseline regression
	(0.1455)
_IafrXnew_g_1	35.1637***
	(1.0889)
N	32367
Chi_ln_pd	1580.0943
p_ln_pd	0.0000
Chi_ln_pop	12780.4619
p_ln_pop	0.0000
Chi_gdp_gr	9.5544
p_gdp_gr	0.0020
Chi_inf	8281.7145
p_inf	0.0000
Chi_kkm	9608.6463
p_kkm	0.0000
Chi_ca_gdp	4513.9768
p_ca_gdp	0.0000
Chi_new_geobrpen	1051.8437
p_new_geobrpen	0.0000
Chi_new_geobrpen	1051.8437
p_new_geobrpen	0.0000


Any idea would be appreciated.

*
*   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/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index