On Tue, 11 Oct 2011, Akram Khaleghei Ghosheh balagh wrote:
Hello ;
I am doing a regression of count data (number of award and there are some
covariates)
I have estiamted the parameters of negative binomial distribuion (lambda is
a function of covaraites, GLM model) by glm.nb function and training
dataset.
Now I want to predict the number of award (for example y=0, y=1, y=2,) or
testing dataset. I dont know how to calculate this numbers?
Do you want the expected probabilities of observing 0, 1, 2, etc.? Then
you can use the convenience function predprob() in package "pscl". For
example:
## package and data
library("pscl")
data("bioChemists", package = "pscl")
## NB model
m <- glm.nb(art ~ ., data = bioChemists)
## predicted probabilities for counts 0, 1, ..., 19 per observation
p <- predprob(m)
## sums across all observations
colSums(p)
which yields:
0 1 2 3 4 5
277.7900354 249.1239076 164.7450252 97.2999696 54.7265588 30.2456638
6 7 8 9 10 11
16.7394187 9.4002205 5.4081462 3.2096223 1.9731828 1.2584834
12 13 14 15 16 17
0.8320010 0.5686597 0.4003506 0.2891672 0.2134580 0.1604948
18 19
0.1225642 0.0948470
hth,
Z
I would be very grateful if anybody could help me.
thanks
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
