Wild Hog Rooting 
179 
was not a full rank and therefore could not be properly analyzed 
by PROC DISCRIM. 
PROC DISCRIM revealed that the 1990 data also failed to 
meet the assumption of equality between within-group covariance 
matrices (P = 0.0008). As was done for 1989 data, we ran PROC 
DISCRIM with a pooled covariance matrix. Results of this classifi- 
cation were almost as good as in 1989. Of the 139 unsuccessful 
traps, 61.0% were correctly classified. The Kappa statistic indicated 
that classifications were also significantly different from zero (Kappa 
= 0.26, P < 0.0001). Standardized coefficients (Table 4) indicated 
that trap success was directly related to percent exposed rock and 
inversely related to percent canopy cover, percent total herb cover, 
and distance to closest understory tree. These relationships are simi- 
lar to the previous year’s except that in 1989 percent canopy cover 
was directly related to trap success. The group centroid on the 
discriminant axis for successful traps (0.36) was significantly differ- 
ent (F = 3.93, P - 0.004) from the centroid for unsuccessful traps 
(-0.20). 
None of the 1990 variables was significantly different between 
rooted and unrooted sites (one-way ANOVAs, Table 5). 
DISCUSSION 
Mammals: Population Trends 
Our findings are in agreement with Singer et al. (1984) that 
there was no significant difference in populations of deer mice be- 
tween rooted and unrooted sites. We believe that the semi-arboreal 
habits (Wolff and Hurlbutt 1982, Singer et al. 1984), choice of 
food items (Howe et al. 1981, Linzey and Linzey 1973), and gen- 
eralist nature (Baker 1968) of deer mice allow them to quickly 
adapt to or not be affected by hog rooting disturbances. Although 
we did not capture enough shrews or voles to allow a quantitative 
comparison, their presence on rooted sites indicates that populations 
of these mammals are not permanently extirpated from rooted areas 
and that they may recolonize disturbed areas quickly. 
Vegetation: Microhabitat Selection 
One problem all habitat studies of this type face is the 
assumption that the reason a particular habitat is unused is that the 
habitat in that area is unsuitable for the animal (Johnson 1981, 
James and McCulloch 1990). However, habitat might not be used 
simply because of low population size. Our classification rates were 
probably lowered by this circumstance because mouse populations at 
some sites were extremely low. 
