parfm: Parametric Frailty Models in R

Frailty models are getting more and more popular to account for overdispersion and/or clustering in survival data. When the form of the baseline hazard is somehow known in advance, the parametric estimation approach can be used advantageously. Nonetheless, there is no unified widely available software that deals with the parametric frailty model. The new parfm package remedies that lack by providing a wide range of parametric frailty models in R. The gamma, inverse Gaussian, and positive stable frailty distributions can be specified, together with five different baseline hazards. Parameter estimation is done by maximising the marginal log-likelihood, with right-censored and possibly left-truncated data. In the multivariate setting, the inverse Gaussian may encounter numerical difficulties with a huge number of events in at least one cluster. The positive stable model shows analogous difficulties but an ad-hoc solution is implemented, whereas the gamma model is very resistant due to the simplicity of its Laplace transform.

• metadata
• relations

Document type	Article de périodique (Journal article) – Article de recherche
Access type	Accès restreint
Publication date	2012
Language	Anglais
Journal information	"Journal of Statistical Software" - Vol. 51, no.11 (2012)
Peer reviewed	yes
Publisher	American Statistical Association ((United States) [S.l.])
issn	1548-7660
e-issn	1548-7660
Publication status	Publié
Affiliation	UCL - SSH/LIDAM/ISBA - Institut de Statistique, Biostatistique et Sciences Actuarielles
Keywords	parametric frailty models ; lognormal ; R ; parfm ; survival analysis ; gamma ; positive stable ; inverse gaussian ; weibull ; exponential ; gompertz ; loglogistic
Links	http://hdl.handle.net/2078.1/135889[Handle]

See also
	[boreal:110123]	parfm: Parametric Frailty Models in R