[18888] in s-news-athena
[S] SURVIVAL QUESTION
daemon@ATHENA.MIT.EDU (Jaime Gomez)
Sun Aug 8 15:57:11 1999
Message-Id: <19990808192418.68286.qmail@hotmail.com>
From: "Jaime Gomez" <eseplus@hotmail.com>
To: s-news@wubios.wustl.edu
Date: Sun, 08 Aug 1999 21:24:17 CEST
Mime-Version: 1.0
Content-Type: text/plain; format=flowed
Dear S-Plus users,
I am estimating the probability that an entity enters a market in a given
year through the use of survival models. To do this, I have data on the
evolution of several companies and several markets on a year basis. Given
that not many ties are present I have estimated three models in order to see
if the results of a continuous and a discrete model were very different.
First, following the counting processes formulation I estimated a Cox
Proportional Hazards Model. Second, I followed Cox (1972) suggestions and I
estimated a logit model over my data. Third, the Prentice and Gloeckler
approximation was used. As it can be seen, the results of the two first
estimations are fairly close. However, the same does not happen with the
third one. What surprises me is the high and different value of the t-ratios
in the third model. Is this normal?. Am I doing anything wrong?.
Thanks a lot for your help.
Jaime Gomez
University of Zaragoza.
SPAIN.
(1) ANDERSEN AND GILL MODEL
*** Cox Proportional Hazards ***
Call:
coxph(formula = Surv(ANO, TCIERRE, CENSURA) ~ TAM1000 + NM + INTERA +
MARGEND + PROXIM1 +
cmobc + HBYC + COMPOT1 + INTDEM + DENSP.INE. + cmbc3, data = estimac2,
na.action
= na.omit, eps = 0.0001, iter.max = 10, method = "efron", robust = F)
n= 27954
coef exp(coef) se(coef) z p
TAM1000 0.026931 1.027297 0.001637 16.450 0.00000
NM 0.161027 1.174716 0.046612 3.455 0.00055
INTERA 2.886710 17.934213 1.490344 1.937 0.05300
MARGEND -0.027478 0.972896 0.154391 -0.178 0.86000
PROXIM1 1.863983 6.449374 0.190187 9.801 0.00000
cmobc -8.227791 0.000267 3.048055 -2.699 0.00690
HBYC -2.539006 0.078945 1.857741 -1.367 0.17000
COMPOT1 -0.095610 0.908818 0.028313 -3.377 0.00073
INTDEM -0.420822 0.656507 0.588315 -0.715 0.47000
DENSP.INE. -0.000573 0.999427 0.000714 -0.804 0.42000
cmbc3 -0.004894 0.995118 0.026184 -0.187 0.85000
(2) LOGIT MODEL.
*** Generalized Linear Model ***
Call: glm(formula = CENSURA ~ TAM1000 + NM + INTERA + MARGEND + PROXIM1 +
cmobc + HBYC +
COMPOT1 + INTDEM + DENSP.INE. + cmbc3, family = binomial(link = logit),
data =
estimac2, na.action = na.omit, control = list(epsilon = 0.001, maxit = 50,
trace
= F))
Deviance Residuals:
Min 1Q Median 3Q Max
-2.402747 -0.06774497 -0.05067618 -0.03864733 3.645177
Coefficients:
Value Std. Error t value
(Intercept) -5.0465577580 0.5611458078 -8.9933092
TAM1000 0.0342797920 0.0022860379 14.9952860
NM 0.1073859754 0.0526272617 2.0405009
INTERA 1.7577232775 1.5122973509 1.1622868
MARGEND 0.0704343396 0.1659165139 0.4245168
PROXIM1 2.2657006140 0.2145488626 10.5603012
cmobc -8.5462113518 3.0091541413 -2.8400710
HBYC -4.9561367096 2.2758560458 -2.1777022
COMPOT1 -0.0939520923 0.0289615791 -3.2440252
INTDEM -0.4335673623 0.5405667747 -0.8020607
DENSP.INE. -0.0008537292 0.0008542989 -0.9993331
cmbc3 0.0313416677 0.0241968072 1.2952811
(3) PRENTICE AND GLOECKLER MODEL
*** Generalized Linear Model ***
Call: glm(formula = CENSURA ~ TAM1000 + NM + INTERA + MARGEND + PROXIM1 +
cmobc + HBYC +
COMPOT1 + INTDEM + DENSP.INE. + cmbc3, family = binomial(link = cloglog),
data =
estimac2, weights = NBRISK2, na.action = na.omit, control = list(epsilon =
0.001,
maxit = 50, trace = F))
Deviance Residuals:
Min 1Q Median 3Q Max
-129.752 -3.968318 -2.999257 -2.314944 213.456
Coefficients:
Value Std. Error t value
(Intercept) -4.6959729546 0.00982136183 -478.13868
TAM1000 0.0278105965 0.00003362837 826.99816
NM 0.1478389427 0.00088689042 166.69358
INTERA 0.9902200141 0.02797453371 35.39719
MARGEND -0.0066270394 0.00270795318 -2.44725
PROXIM1 1.7957904288 0.00368347246 487.52650
cmobc -9.4695313722 0.05603265877 -169.00021
HBYC -2.7854023239 0.03780347623 -73.68112
COMPOT1 -0.0734599056 0.00046771642 -157.06078
INTDEM -1.1773251877 0.00971900918 -121.13634
DENSP.INE. -0.0001575185 0.00001414123 -11.13895
cmbc3 0.0371676247 0.00038871998 95.61542
______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
