[R] Shafer's MI software for S-plus

[(Ted Harding)]

On 01-Apr-03 Ted Harding wrote:
> On 01-Apr-03 Prof Brian Ripley wrote:
>> I have an R port of MIX, but it is not 100% reliable. However, neither
>> is the S-PLUS original!  (We have found several bugs already.)
>> 
>> When I have a few spare days (if ever?) I will try again.  We got
>> enough to work for the project we needed it for.  Meanwhile,
>> 
>> http://www.stats.ox.ac.uk/pub/bdr/mix_1.0-1.tar.gz
>> 
>> has the current state.
> 
> Thanks, Brian! I'll see how it goes on the somewhat recalcitrant
> dataset I'm looking at, and report back on anything interesting.

Got it installed in R (version 1.6.2 on SuSE Linux 7.2). Apparently
trouble free.

Worked through the case-study "stlouis.dat" according to Shafer's
example in the README. All went well until the final batch of
commands below:

   # Now try some imputations. The following commands produce three
   # multiple imputations under the alternative model. The imputations
   # are proper if we can assume that the data augmentation procedure
   # achieves stationarity by 100 steps.
   rngseed(454545)
** newtheta<-dabipf.mix(s,margins,design,thetahat4,steps=100,showits=T)
** imp1<-imp.mix(s,newtheta,x)
** newtheta<-dabipf.mix(s,margins,design,newtheta,steps=100,showits=T)
** imp2<-imp.mix(s,newtheta,x)
   newtheta<-dabipf.mix(s,margins,design,newtheta,steps=100,showits=T)
   imp3<-imp.mix(s,newtheta,x)

> newtheta<-dabipf.mix(s,margins,design,newtheta,steps=100,showits=T)
Steps of Data Augmentation-Bayesian IPF: 
1...2...3...4...5...6...7...8...9...10...11...12...13...14...15...
16...17...18...19...20...21...22...23...24...25...26...27...28...
Segmentation fault

Or it may segfault on the imp1 step or the imp2 step.

Is this one of the "mix" bugs, or some R/Linux version incompatibility?
A memory problem (though plating a trace-print of "gc()" indicates that
there's plenty of R memory to spare)?

Any comments welcome.

With thanks,
Ted.


--------------------------------------------------------------------
E-Mail: (Ted Harding) <Ted.Harding at nessie.mcc.ac.uk>
Fax-to-email: +44 (0)870 167 1972
Date: 01-Apr-03                                       Time: 19:08:56
------------------------------ XFMail ------------------------------