An Illustration of the Use of Generalized Linear Models to Measure Long-Term Trends in the Wet Deposition of Sulfate (1994)

Abstract:

In this research, I present a method to measure long-term trends in the wet deposition of sulfate, adjusting for effects of season and meteorology. The method­ology proposed incorporates the use of generalized linear models, specifically gamma regression models, which are a useful extension of previous efforts applying ordinary least squares regression models to precipitation monitoring data; Gamma regression models are appropriate for right-skewed, positive data and alleviate the problems introduced by fitting regression models to such data on a transformed scale. For the application presented here, the gamma regression models provide simple estimates of long-term trends in the wet deposition of sulfate. While these trend estimates are very similar to estimates produced by ordinary least squares regression models fitted to the log-transformed data, I discuss other applications where it is more advantageous to fit regression models on the untransformed scale. 

Keywords:

gamma regression, environmental monitoring, lognormal distribution