[R] Standard error of quasi-Poisson regression coefficients in mgcv

YuHao yuhao7510 at hotmail.com
Sat Jan 24 23:00:55 CET 2015


Hi, I am analyzing the relationship between animal density and several environmental factors using GAM in mgcv. I used Poisson distribution and quasi-Poisson model separately because there is overdispersion. In my model, there is a categorical predictor variable "Year". However, when I compared the outputs from Poisson and quasi-Poisson models, I realized that the standard errors of "Year" from quasi-Poisson model were even a little smaller than those from the Poisson model. Based on my knowledge, the standard errors of quasi-Poisson regression coefficients should be larger than those from Poisson model due to dispersion. Could anybody help me figure out what I was doing wrong? Thank you very much!Hao
  		 	   		  
	[[alternative HTML version deleted]]



More information about the R-help mailing list