Dear all, I am trying to find a suitable R-function for parametric proportional hazard regressions. The package survival contains the coxph() function which performs a Cox regression which leaves the base hazard unspecified, i.e. it is a semi-parametric method. The package Design contains the function pphsm() which is good for parametric proportional hazard regressions when the underlying base distribution is "weibull" or "exponential". But what if I need a parametric proportional hazard model with the other "usual" distributions like '"gaussian"', '"logistic"', '"lognormal"'and '"loglogistic"? Thanks a lot for your support! Valentin Dimitrov Statistics and Econometrics University of Saarland
On Fri, 7 Jul 2006, Valentin Dimitrov wrote:> I am trying to find a suitable R-function for > parametric proportional hazard regressions. The > package survival contains the coxph() function which > performs a Cox regression which leaves the base hazard > unspecified, i.e. it is a semi-parametric method. The > package Design contains the function pphsm() which is > good for parametric proportional hazard regressions > when the underlying base distribution is "weibull" or > "exponential". But what if I need a parametric > proportional hazard model with the other "usual" > distributions like '"gaussian"', '"logistic"', > '"lognormal"'and '"loglogistic"?Those are not proportional hazards families of distributions. That is, if the distribution is gaussian for one value of the hazard ratio parameters it will not be gaussian for any other value. You can get accelerated failure models with these distributions using survreg() in the survival package. -thomas Thomas Lumley Assoc. Professor, Biostatistics tlumley at u.washington.edu University of Washington, Seattle
Valentin, Have you tried survreg() in the Design library? Regards, -Cody -----Original Message----- From: r-help-bounces at stat.math.ethz.ch [mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of Valentin Dimitrov Sent: Friday, July 07, 2006 10:1 AM To: r-help at stat.math.ethz.ch Subject: [R] parametric proportional hazard regression Dear all, I am trying to find a suitable R-function for parametric proportional hazard regressions. The package survival contains the coxph() function which performs a Cox regression which leaves the base hazard unspecified, i.e. it is a semi-parametric method. The package Design contains the function pphsm() which is good for parametric proportional hazard regressions when the underlying base distribution is "weibull" or "exponential". But what if I need a parametric proportional hazard model with the other "usual" distributions like '"gaussian"', '"logistic"', '"lognormal"'and '"loglogistic"? Thanks a lot for your support! Valentin Dimitrov Statistics and Econometrics University of Saarland ______________________________________________ R-help at stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html This e-mail, facsimile, or letter and any files or attachments transmitted with it contains information that is confidential and privileged. This information is intended only for the use of the individual(s) and entity(ies) to whom it is addressed. If you are the intended recipient, further disclosures are prohibited without proper authorization. If you are not the intended recipient, any disclosure, copying, printing, or use of this information is strictly prohibited and possibly a violation of federal or state law and regulations. If you have received this information in error, please notify Baylor Health Care System immediately at 1-866-402-1661 or via e-mail at privacy at baylorhealth.edu. Baylor Health Care System, its subsidiaries, and affiliates hereby claim all applicable privileges related to this information.
> > Those are not proportional hazards families of > distributions. That is, if > the distribution is gaussian for one value of the > hazard ratio parameters > it will not be gaussian for any other value. > > You can get accelerated failure models with these > distributions using > survreg() in the survival package. > > > -thomasI do not need a accelerated failure model, but a proportional hazard model with a f0= weibull, exponential, loglogistic or lognormal baseline distribution. The hazard function is lambda(t)=exp(Xi*beta)*lambda0(t), where lambda0 is the baseline hazard lambda0(t)=f0(t)/(1-F0(t)) where f0 and F0 are the baseline density and cumulative distribution functions. This is a proportional hazard model since the ratio lambda(t|Xi)/lambda(t|Xj)=exp(Xi*beta)/exp(Xj*beta) does not depend on t. Valentin
Cody, I have tried the survreg() in the Design library, it is analogous to survreg() in the survival library and it seems to me it is designed only for accelerated time models like the accelerated failure model (or accelerated lifetime model) and not for proportional hazard models. Correct me if I am wrong. Valentin --- "Hamilton, Cody" <CodyH at BaylorHealth.edu> wrote:> > Valentin, > > Have you tried survreg() in the Design library? > > Regards, > -Cody >
Valentin Dimitrov wrote:> Cody, > > I have tried the survreg() in the Design library, it > is analogous to survreg() in the survival library and > it seems to me it is designed only for accelerated > time models like the accelerated failure model (or > accelerated lifetime model) and not for proportional > hazard models. Correct me if I am wrong. > > Valentinsurvreg does not exist in the Design package. You may be thinking of psm which is a wrapper function for survreg.> > --- "Hamilton, Cody" <CodyH at BaylorHealth.edu> wrote: > >> Valentin, >> >> Have you tried survreg() in the Design library? >> >> Regards, >> -Cody >> > > ______________________________________________ > R-help at stat.math.ethz.ch mailing list > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html >-- Frank E Harrell Jr Professor and Chair School of Medicine Department of Biostatistics Vanderbilt University
On Fri, 7 Jul 2006, Valentin Dimitrov wrote:> > I do not need a accelerated failure model, but a > proportional hazard model with a f0= weibull, > exponential, loglogistic or lognormal baseline > distribution. The hazard function is > lambda(t)=exp(Xi*beta)*lambda0(t), > where lambda0 is the baseline hazard > lambda0(t)=f0(t)/(1-F0(t)) where f0 and F0 are the > baseline density and cumulative distribution > functions. > This is a proportional hazard model since the ratio > lambda(t|Xi)/lambda(t|Xj)=exp(Xi*beta)/exp(Xj*beta) > does not depend on t. >For a weibull (including exponential) model you can do this with survreg. For the other models you would have to maximize the likelihood directly. This will involve writing the likelihood directly in terms of the hazard and cumulative hazard, since a proportional hazards model that is gaussian at X=0 is not gaussian at any other X. -thomas Thomas Lumley Assoc. Professor, Biostatistics tlumley at u.washington.edu University of Washington, Seattle