Application of Regression Models in Bird Population Data: An Example of Haçlı Lake

Abstract

In this study, the effects of habitat, ordo, UTM frame, seasons and number of species on bird populations and distribution in Haçlı Lake were investigated. Bird population data were obtained using point counts and transect observation methods. Poisson regression is typically used in such data sets. The basic principle of Poisson regression assumes that the variance is equal to the mean. Failure to achieve this equality causes incorrect parameter estimates and standard errors. In practice, the variance is often higher than the mean (variance > mean). This is called over-dispersion, where the value of over-dispersion is greater than 1.0. The population status of the data set used in the study was over-dispersed. Negative binomial regression is the most common method used to eliminate the over-dispersion effect. In this case, the preferred method is the negative binomial regression method. The over-dispersion value in the Poisson regression was considerably greater than 1.0 (54.937) while the over-dispersion value was very close to 1.0 (1.588) in the negative binomial regression. The results indicated that the use of negative binomial regression method is more appropriate. Therefore, parameter estimations were interpreted according to negative binomial regression method. Herein, climatic factors including temperature and humidity exhibited significant impacts on population density and number of species.In this study, the effects of habitat, ordo, UTM frame, seasons and number of species on bird populations and distribution in Haçlı Lake were investigated. Bird population data were obtained using point counts and transect observation methods. Poisson regression is typically used in such data sets. The basic principle of Poisson regression assumes that the variance is equal to the mean. Failure to achieve this equality causes incorrect parameter estimates and standard errors. In practice, the variance is often higher than the mean (variance > mean). This is called over-dispersion, where the value of over-dispersion is greater than 1.0. The population status of the data set used in the study was over-dispersed. Negative binomial regression is the most common method used to eliminate the over-dispersion effect. In this case, the preferred method is the negative binomial regression method. The over-dispersion value in the Poisson regression was considerably greater than 1.0 (54.937) while the over-dispersion value was very close to 1.0 (1.588) in the negative binomial regression. The results indicated that the use of negative binomial regression method is more appropriate. Therefore, parameter estimations were interpreted according to negative binomial regression method. Herein, climatic factors including temperature and humidity exhibited significant impacts on population density and number of species.