Constant in Cox regression

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

Constant in Cox regression

Hector Maletta

One student has posed to me a question I could not answer. The Cox regression survival model is based on the hazard function h(t)= h0(t) x EXP(b0 + b1 X1 + b2 X2 +……+bk Xk), where h0(t) is the baseline hazard function, dependent only on time, and the exponential EXP(.) is a function of covariates. Now, SPSS output for Cox regression includes the baseline hazard function h0(t) and also the b coefficients for all covariates X, EXCEPT for the constant intercept b0, the first term in the logit. It is nowhere in the output, as far as I can see. May somebody explain how does one get that constant?

Thanks for any help.

 

Hector

Reply | Threaded
Open this post in threaded view
|

Re: Constant in Cox regression

Alex Reutter

h(t)= h0(t) x EXP(b0 + b1 X1 + b2 X2 +……+bk Xk)
  = h0(t) x EXP(b0) x EXP(b1 X1 + b2 X2 +……+bk Xk)
  = h*0(t) x EXP(b1 X1 + b2 X2 +……+bk Xk)

where h*0(t) = h0(t) x EXP(b0).  

It is not possible to estimate the Cox Regression model as you have written it, because h0(t) and EXP(b0) cannot be estimated separately; you can arbitrarily rescale the estimate of h0(t) by a factor "a" by multiplying the estimate of EXP(b0) by "1/a".  In other words, the baseline hazard function is the Cox regression equivalent of the intercept.

Alex



From: Hector Maletta <[hidden email]>
To: [hidden email]
Date: 10/08/2010 09:02 PM
Subject: Constant in Cox regression
Sent by: "SPSSX(r) Discussion" <[hidden email]>





One student has posed to me a question I could not answer. The Cox regression survival model is based on the hazard function h(t)= h0(t) x EXP(b0 + b1 X1 + b2 X2 +……+bk Xk), where h0(t) is the baseline hazard function, dependent only on time, and the exponential EXP(.) is a function of covariates. Now, SPSS output for Cox regression includes the baseline hazard function h0(t) and also the b coefficients for all covariates X, EXCEPT for the constant intercept b0, the first term in the logit. It is nowhere in the output, as far as I can see. May somebody explain how does one get that constant?
Thanks for any help.
 
Hector

Reply | Threaded
Open this post in threaded view
|

Re: Constant in Cox regression

Hector Maletta

Thanks, Alex.

 

Hector

 

De: SPSSX(r) Discussion [mailto:[hidden email]] En nombre de Alex Reutter
Enviado el: Saturday, October 09, 2010 1:12 AM
Para: [hidden email]
Asunto: Re: Constant in Cox regression

 


h(t)= h0(t) x EXP(b0 + b1 X1 + b2 X2 +……+bk Xk)
  = h0(t) x EXP(b0) x EXP(b1 X1 + b2 X2 +……+bk Xk)
  = h*0(t) x EXP(b1 X1 + b2 X2 +……+bk Xk)

where h*0(t) = h0(t) x EXP(b0).  

It is not possible to estimate the Cox Regression model as you have written it, because h0(t) and EXP(b0) cannot be estimated separately; you can arbitrarily rescale the estimate of h0(t) by a factor "a" by multiplying the estimate of EXP(b0) by "1/a".  In other words, the baseline hazard function is the Cox regression equivalent of the intercept.

Alex


From:

Hector Maletta <[hidden email]>

To:

[hidden email]

Date:

10/08/2010 09:02 PM

Subject:

Constant in Cox regression

Sent by:

"SPSSX(r) Discussion" <[hidden email]>

 





One student has posed to me a question I could not answer. The Cox regression survival model is based on the hazard function h(t)= h0(t) x EXP(b0 + b1 X1 + b2 X2 +……+bk Xk), where h0(t) is the baseline hazard function, dependent only on time, and the exponential EXP(.) is a function of covariates. Now, SPSS output for Cox regression includes the baseline hazard function h0(t) and also the b coefficients for all covariates X, EXCEPT for the constant intercept b0, the first term in the logit. It is nowhere in the output, as far as I can see. May somebody explain how does one get that constant?
Thanks for any help.
 
Hector

Se certificó que el correo entrante no contiene virus.
Comprobada por AVG - www.avg.es
Versión: 8.5.448 / Base de datos de virus: 271.1.1/3185 - Fecha de la versión: 10/08/10 18:34:00