Correct Interpretation of survreg() coeffs

[wwreith]

Am I correct in assuming that the output below essentially translates to "Males have a mean time that is significantly lower than Females"? Is this the correct way to interpret the fact that the coefficient is negative? Assume the variale sex is treated as a factor with Female =0 and Male=1.

survmodel<-survreg(survobj~sex,data=data1, dist="weibull")
survsum<-summary(survmodel)
survsum
                            Value      Std. Error     z           p
SexMale        -0.47830     0.3745 -1.2770 2.02e-01

[David Winsemius]

On May 4, 2012, at 9:43 AM, wwreith wrote:

> Am I correct in assuming that the output below essentially  
> translates to
> "Males have a mean time that is significantly lower than Females"?  
> Is this
> the correct way to interpret the fact that the coefficient is  
> negative?

  I wouldn't be using exactly that phrasing since I do not think AFT  
models are estimating means. I might say the  "point estimate for the  
time to event for Sex==Male is lower than for the Female". You should  
plot the predicted survival curves. I also do not think the  
inferential statistics allow you to say that the difference is  
"significantly different". Your z-stat is only -1.28 and your p-value  
is only 0.202.

> Assume the variale sex is treated as a factor with Female =0 and  
> Male=1.
>
> survmodel<-survreg(survobj~sex,data=data1, dist="weibull")
> survsum<-summary(survmodel)
> survsum
>                            Value      Std. Error     z           p
> SexMale        -0.47830     0.3745 -1.2770 2.02e-01
>
> --
> View this message in context: http://r.789695.n4.nabble.com/Correct-Interpretation-of-survreg-coeffs-tp4608655.html

--

David Winsemius, MD
Heritage Laboratories
West Hartford, CT

______________________________________________
[hidden email] 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.