Displaying 1 result from an estimated 1 matches for "anomaly1".
Did you mean:
anomaly
2010 Dec 07
0
coxph failure
...ph. I'll have to think about what
to do since on the one hand it's your own fault for trying to fit a very
bad model, and on the other I'd like the routine to give a nice error
message before it dies.
In the data set you sent me the predictor variable is very skewed:
> quantile(anomaly1$CREAT, c(0, .5, .9, .999, 1))
0% 50% 90% 99.9% 100%
0.5000 1.4000 2.2000 4.6027 7.3000
There are 1000 observations, the largest is 7.3 and the second largest
is 4.6. The subject with a 7.3 was a failure. When you fit a 4 degree
of freedom polynomial to this there is enough "s...