Zero-inflated regression models: predicting no 0s
On Thu, 2 Jun 2011, geojs wrote:
Thanks for the quick reply, I understand that the predict(zip1A, type = "response") command is computing the fitted_means and these are different than the probabilities predict(zip1A, type = "prob").
Yes. One evaluates the probability density function, and the other one the expectation from this density.
Although, according to Martin (2005), the highest probabilities do not simply lead to the true count estimates: "to get the true estimate of relative mean abundance from the ZIP one must multiply the estimated relative mean number of individuals at a site by the probability that the relative mean number of individuals at a site is generated through a Poisson distribution."
I haven't checked that paper but I suspect that this is a verbal description of the fitted mean of a zero-inflated Poisson distribution. See Equation 8 in the JSS paper that introduces hurdle/zeroinfl.
I initially thought that the predicted mean and the observed count could be compared to estimate the fit of the model,
They can be. But then - not surprisingly - you only assess the fit of the mean. (E.g., you do not assess the ability to predict the number of zeros, and you do not assess potential overdispersion etc.) Z
but now I am not sure what to think with Martin (2005) statement. Thank you for your help, JM Martin, T.G. et al. (2005) Zero tolerance ecology: improving ecological inference by modelling the source of zero observations, Ecology Letters, Volume 8, Issue 11, pages 1235?1246. -- View this message in context: http://r.789695.n4.nabble.com/Zero-inflated-regression-models-predicting-no-0s-tp3564807p3568865.html Sent from the R help mailing list archive at Nabble.com.
______________________________________________ R-help at 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.