Abstract

Methods for analyzing over-dispersed count data in a one-way layout were compared using a Monte Carlo study. Several variance stabilizing transformations were examined as alternatives to analyzing the raw data using a general linear model. Additionally, generalized linear models were fit using a log link. For the generalized linear model, three approaches to account for over-dispersion were investigated: (1) a negative binomial distribution with known k, (2) a Poisson distribution with Pearson's X2 as an estimate of the scale parameter, and (3) a Poisson distribution with over-dispersion estimated using the deviance. The analysis of the raw data and log transformed data controlled the size of the tests better than the generalized linear models in the region of the sample space studied.

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Download

Share

COinS
 
Apr 26th, 10:00 AM

THE ANALYSIS OF OVER-DISPERSED COUNT DATA FROM A SINGLE FACTOR STUDY

Methods for analyzing over-dispersed count data in a one-way layout were compared using a Monte Carlo study. Several variance stabilizing transformations were examined as alternatives to analyzing the raw data using a general linear model. Additionally, generalized linear models were fit using a log link. For the generalized linear model, three approaches to account for over-dispersion were investigated: (1) a negative binomial distribution with known k, (2) a Poisson distribution with Pearson's X2 as an estimate of the scale parameter, and (3) a Poisson distribution with over-dispersion estimated using the deviance. The analysis of the raw data and log transformed data controlled the size of the tests better than the generalized linear models in the region of the sample space studied.