[R-sig-Geo] Binom.krige.bayes sim.predict question

Nicola Batchelor N.A.Batchelor at sms.ed.ac.uk
Wed Jul 29 11:26:13 CEST 2009


Dear list,

I've been using geoRglm  for parameter estimation and spatial predictions
for a disease dataset along with 3 predictor variables.  

I just have a quick question about the option sim.predict in
output.glm.control - am I correct in thinking that when this is set to TRUE,
you end up with a sample from the predictive distribution at each prediction
location, and from this sample the predictive median, 95%CrI and uncertainty
is calculated for each prediction location?

If that is correct, then how are predictive medians calculated when this
option is set to FALSE?  I have run my model like this and ended up with a
sensible looking prediction, but if I am correct in the above understanding
of sim.predict, then how are the medians calculated without the sample from
the predictive distributions?  

If Im not correct, can someone please point me in the right direction as to
how the predictive output is produced when this is TRUE and FALSE?

Many thanks (and I hope this question makes sense!)


Nicola


-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.



More information about the R-sig-Geo mailing list