Question about zero-inflated Poisson glmer
Dear Philipp, I'm wondering if there is some kind of 'detection limit' in the dataset. After looking at the data I get the feeling that all values below 10 are set at zero. The relation between the counts and the covariates a and b are not linear. They have an optimum. The proportion of zero's is not constant but varies with the covariates. So you need to model that can handle that. Many models assume that the zero-inflation is constant. I'd settle for a set of two models: a logistic regression for the zero or count and a negative binomial for the non-zero counts. A truncate negative binomial distribution (that doesn't have values below 10) would be ideal. Best regards, ir. Thierry Onkelinx Instituut voor natuur- en bosonderzoek / Research Institute for Nature and Forest team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance Kliniekstraat 25 1070 Anderlecht Belgium To call in the statistician after the experiment is done may be no more than asking him to perform a post-mortem examination: he may be able to say what the experiment died of. ~ Sir Ronald Aylmer Fisher The plural of anecdote is not data. ~ Roger Brinner The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data. ~ John Tukey 2016-06-28 10:51 GMT+02:00 David Duffy <David.Duffy at qimrberghofer.edu.au>:
On Tue, 28 Jun 2016, Philipp Singer wrote: You can find a sample of the data here:
https://www.dropbox.com/s/kqqxmc3wp225lug/r_sample.csv.gz?dl=1 You can think of the setting as popularity of items "y" inside stores "id" explained by two features "a" and "b" whereas "a" is more of a control covariate and I am interested in whether "b" has a positive impact.
Probably not very helpful (what a horrible distribution y has), but library(ordinal) x$cats <- cut(x$y, c(-1,0,10,100,1000,20000)) c1 <- clmm2(cats ~ a + (1|id), data=x) c2 <- clmm2(cats ~ a + b + (1|id), data=x) does run... Cheers, David Duffy.
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models