s-news
[Top] [All Lists]

Initial values for parameters when using censorReg

To: s-news@wubios.wustl.edu
Subject: Initial values for parameters when using censorReg
From: Denise <dbabinea@math.uwaterloo.ca>
Date: Fri, 24 Oct 2003 16:43:47 -0400 (EDT)
Hi,

I am trying to use censorReg to fit an exponential model to data that
is left, right and interval censored. Here is the command that I use.

fit<-censorReg(censor(data$lower,data$upper,data$status)~1,weights=data$weight
,distribution="exponential")

For some reason, I get these warning messages:

Warning messages:
1: problems in covariance matrix computations; error code iervcv = 1
in: censorReg.mlest(Yi, censor.codesi, Xi, casewti, distribution, link,
....
2: problems in covariance matrix computations; error code iervcv = 1
in: censorReg.mlest(Yi, censor.codesi, NULL, casewti, distribution, link,
....

I next plotted the likelihood for my data under an exponential model to
see what initial value I could use in censorReg to find the m.l.e..
I wasn't sure if this was the problem but it was worth trying. It turns
out that the likelihood could be maximized so I just put in my value,
say init.val, into censorReg. This is how I did it.

fit<-censorReg(censor(data$lower,data$upper,data$status)~1,weights=data$weight
,initial=init.val,distribution="exponential")

This command doesn't do anything different from before so I am confused.
From my plot of the likelihood, I can get the m.l.e. but when I put it
into censorReg..even when I give censorReg a 'good' initial value, I still
get the same error.

Any help would be greatly appreciated.

Denise




<Prev in Thread] Current Thread [Next in Thread>
  • Initial values for parameters when using censorReg, Denise <=