Skip to content

1 continuous non-normal variable ~ 4 factors + 1 continuous covariate (with interactions)

5 messages · Dr. Pablo E. Verde, Alal

#
Hello

Im struggling on something... I have one continuous variable (A), and I need
to explain it with 4 factors, and maybe one continuous covariate. 
And of course, my variable A is not normal at all (it's a duration in
seconds, whole numbers). 

What can I do? I would know how do deal with it if I had one factor, maybe
two, but in that case Im really not sure what I am supposed to do to write a
model that is statistically correct.

Thanks,

Al

--
View this message in context: http://r.789695.n4.nabble.com/1-continuous-non-normal-variable-4-factors-1-continuous-covariate-with-interactions-tp3444378p3444378.html
Sent from the R help mailing list archive at Nabble.com.
#
If you are working with a positive outcome variable, one approach is to use 
glm( ..., family ="Gamma"),
other approach is to use survreg() in the package survival.

Cheers,

Pablo
#
Thanks

Im not sure about the gamma, but a survival analysis seems appropriate, but
does it work for factors and continuous covariates? Do you have to verify
some conditions beforehand? 

--
View this message in context: http://r.789695.n4.nabble.com/1-continuous-non-normal-variable-4-factors-1-continuous-covariate-with-interactions-tp3444378p3444670.html
Sent from the R help mailing list archive at Nabble.com.
#
Zitat von Alal <bohk at voila.fr>:
Here is an example:

# test data...
library(survival)

set.seed(1007)
x <- runif(50)
mu <- c(rep(1, 25), rep(2, 25))

test1 <- data.frame(Time = qsurvreg(x, mean = mu, scale= 0.5, distribution =
"weibull"),
                    Status = rbinom(50,1,0.7),
                    f1 = gl(2, 25),
                    f2 = factor(rbinom(50, 1, 0.5)),
                    f3 = factor(rbinom(50, 1, 0.5)),
                    f4 = factor(rbinom(50, 1, 0.5)),
                    z = rnorm(50)
                    )

mod1 <- survreg(Surv(Time, Status) ~  f1*(f2 + f2 + f4) + z, data = test1)
summary(mod1)

Cheers,

Pablo
#
Thanks, I guess I can do that, and it actually seem appropriate for one of my
variable. 

But can you do post-hoc tests on a survival analysis? Use contrasts or
something? 

--
View this message in context: http://r.789695.n4.nabble.com/1-continuous-non-normal-variable-4-factors-1-continuous-covariate-with-interactions-tp3444378p3446782.html
Sent from the R help mailing list archive at Nabble.com.