Dear all,
I am fitting a parametric regression model to survival data using the
flexsurvreg function from the flexsurv package. I am using a Gompertz
distribution (a 2-parameter distribution) to describe the hazard function
and I want to compare two groups. The model is very simple:
flexsurvreg(formula = Surv(day, censored) ~ group, dist = "gompertz")
and the output is also quite simple:
Maximum likelihood estimates:
est L95%
U95%
shape 0.44600 0.40000
0.49200
rate 0.00406 0.00263
0.00628
group B -0.31800 -0.61700
-0.01820
N = 407, Events: 174, Censored: 233
Total time at risk: 3007
Log-likelihood = -482.789, df = 3
AIC = 971.578
Shape and rate are self-explanatory: they are the estimates of the
(Gompertz hazard) distribution. It's the interpretation of the covariate
coefficient estimate that I'm struggling with.
In the flexsurv package help file it is stated that " Covariates are
included through a linear model on the parameter of the distribution which
determines its mean, for example, the "location", "scale" or "rate"
parameter (...)". For the Gompertz distribution, this would be the "rate"
parameter.
Does this mean that the hazard of group B is described by a Gompertz
distribution with the same shape as that of group A but with rate=
0.00406-0.318?
This yields a negative estimate of rate (that is, a negative background
hazard for group B), which doesn't make any sense.
Any clarification on this would be appreciated. Thanks!
Sara Carvalho
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
