I entered calculated data into a one-way ANOVA model blocked for forest remnant in
which average daily relative abundance was the dependent variable, and edge or interior location
was the independent variable. Because I was not interested in the effect of sampling year, I
averaged the relative abundances for each analyzed species and group between both years. I
tested the data for normality with the Ryan-Joiner test, and for equal-variance with the Levine
test. I square-root transformed non-normal and heteroskedastic distributions for individual
species and groups. I used the non-parametric Friedman test to analyze species and groups
unable to meet parametric test assumptions after transformation. Because there were 5 sampled
forest remnants with both edge and interior locations, this resulted in a total of 10 possible forest
remnant locations. In order to deal with non-normality issues due to having too many zeros in the
data, I only statistically analyzed individual groups in each level of analysis if they were present
in at least half (5) of the 10 possible forest remnant locations. An alpha of 0.1 was used for all
statistical tests.
I calculated species richness (edge and interior) per forest remnant and entered it into a
one-way ANOVA model blocked by forest remnant in which number of species was the
dependant variable, and edge or interior location was the independent variable. Similar to the
count data, I averaged species richness data between both years. I tested normality and variance
assumptions as previously described (alpha=0.1).
Lastly, in order to gauge similarities in species assemblages at edges and interiors, I
computed Horn similarity index values between edges and interiors within each remnant. To do
this, I used R Statistical Program, using the Vegan Community Analysis package.