ZIGP-package {ZIGP} | R Documentation |
Adapts GP, ZIP and ZIGP models by Maximum Likelihood estimation. Regression is allowed not only on the mean but also on overdispersion and zero-inflation level. Therefore, three design matrices are allowed.
Package: | ZIGP |
Type: | Package |
Version: | 2.7 |
Date: | 2008-11-14 |
License: | GPL (>= 3) |
Distribution functions in R notation are 'dzigp', 'pzigp', 'qzigp', 'rzigp'.
The most important function is 'wald.test'. This function can be fed with design matrices for the mean, overdispersion and zero-inflation level. It returns an summary-like overview of the covariates together with significance statistics referring to the Wald test. It can be used for the sequential elimination of non-significant effects.
Other useful functions are 'loglikelihood.zigp', which evaluates the loglikelihood function on the given parameter value. Scores can be calculated using 'gradient', the Fisher Information matrix using 'FM'.
Nonnested model comparison can be done using a test proposed by Vuong which is implemented in function 'vuong' or by a test proposed by Clarke using function 'clarke'.
Vinzenz Erhardt
Maintainer: Vinzenz Erhardt <erhardt@ma.tum.de>
Czado, C., Erhardt, V., Min, A., Wagner, S. (2007) Zero-inflated generalized Poisson models with regression effects on the mean, dispersion and zero-inflation level applied to patent outsourcing rates. Statistical Modelling 7 (2), 125-153.
Masterthesis in German: Erhardt, Vinzenz. Verallgemeinerte Poisson und Nullenueberschuss- Regressionsmodelle mit regressiertem Erwartungswert, Dispersions- und Nullenüberschuß-Parameter und eine Anwendung zur Patentmodellierung. ("http://www-m4.ma.tum.de/Diplarb/"), 2006.
Vuong, Q.H. (1989). Likelihood Ratio tests for model selection and nonnested hypotheses. Econometrica 57(2), 307-333.
Clarke, Kevin A. (2007). A Simple Distribution-Free Test for Nonnested Model Selection. Political Analysis 2007 15(3), 347-363.
Schwarz, G. (1978). Estimating the Dimension of a Model. Annals of Statistics 6, 461-464.