I analyzed the data in a three-step process: first, I
compared life history traits for weeds and non-weeds;
second, I compared life history traits for native and exotic
weeds; and third, I compared life history traits for invasive
and non-invasive exotic weeds. In all comparisons, I
analyzed data using a two-way test of independence
(Sokal and Rohlf 1995). If weediness was independent of
the life history trait, the expected proportion for weeds and
non-weeds (or native and exotic weeds -or invasive and
non-invasive weeds) for a given life history attribute
should be the same and they should be equal to
distribution for all species for that attribute. If weediness
was not independent of the life history trait, then plants
with a given life history attribute were more likely to be
weeds than plants without that attribute.
Because data analyses required multiple tests of
independence, I applied a Bonferroni correction to the P
values (Sokal and Rohlf 1995). When the results were
significant, I subdivided the contingency tables to
determine which attributes were significantly different
(Zar 1999). If the results of the analyses are robust, then
the life history trait should be significant for several weeds
lists. I defined overall significance for a life history trait as
a significant difference in the distribution of life history
attributes in at least two weeds lists with the response in
the same direction. If one weeds list indicated a significant
increase and a second list indicated a significant decrease,
then the response of that life history trait was not
considered significant.
To indicate relative importance of a life history trait, I
calculated odds ratios (Sokal and Rohlf 1995) for the
weeds lists with significant differences in the distribution
of life history attributes and reported averaged ratios in the
discussion section.