USGS Home
Contact USGS
Search USGS

The first approach used related the density to the number of years since burn. This simple approach ignores effects due to the different plots and to the various years. Effects of these confounding variables can complicate interpretation of the effects of the explanatory variables of interest. The density of birds varies from plot to plot because of intrinsic differences in habitat. Also, densities vary annually due to climatic variation, regional changes in the population size, and other influences. To eliminate the effects of these "nuisance" variables and focus on the treatment effects, I also took a second approach, which involved two steps. First, I modeled the density of a species as a linear function of plot and year:

In the second step, residuals from that models, termed adjusted densities, were related to the explanatory variable, years since burn. PROC GLM (SAS Institute 1990) was used for the first step. In the second step, the locally weighted regression (loess) procedure of S-PLUS (Statistical Sciences, Inc. 1993) was used. Because there was no variation in the explanatory variable for the control plot (Plot 13), effects were totally confounded with year, so data from that plot were excluded from the analysis of residuals.

The analysis was conducted on 17 of the most common terrestrial species recorded on the plots (Table 1). The American Goldfinch (Carduelis tristis) was omitted because it nests later in the season, so the surveys were inadequate to assess their breeding densities. The Western Kingbird (Tyrannus verticalis), Gray Catbird (Dumetella carolinensis), and swallows were omitted because they nest mostly outside the plots and used the plots only for foraging.