Appendix D. Nonlinear regression analyses modeling the effect of distance from the various classes of remnant vegetation and abiotic factors on the density of naturally regenerating tree and shrub species.
TABLE D1. Summary of results of nonlinear regression analyses, modeling the effect of distance from the various classes of remnant vegetation (biotic) and abiotic factors on the density of naturally regenerating tree and shrub species per 10 m^{2}. See Notes at bottom of table for details.



Forest 
Monocot 
Tree 
Shrub 
Habitat^{††} 
Biotic 
Abiotic 
Total 
F 
P 













Species^{†} 
Ex 
Lin^{‡} 
Ex 
Lin^{§} 
Ex 
Lin^{} 
Ex 
Lin^{¶} 
linear 
linear 
partial 
partial 





















Total density of all individuals 















F[3,87] = 22.65 
<0.0001 
Total number of species 















F[4,86] = 7.24 
0.0002 
1. Winddispersed species (not reproducing) 

Tric ga 













F[3,87] = 6.15 
0.0032 
Lueh se 













F[2,88] = 6.80 
0.0107 
Cord al 













F[1,89] = 17.47 
<0.0001 
2. Winddispersed species (reproducing) 

Coch vi 













F[3,87] = 4.49 
0.0139 
3. Ballisticdispersed species 

Hyba pr 













F[5,85] = 15.85 
<0.0001 
4. Smallbirddispersed species 

Mico ar 













F[2,88] = 9.21 
0.0002 
Cecr in 













F[6,84] = 12.03 
<0.0001 
Bana gu 















Trem mi 















Psyc gr 













F[2,88] = 7.59 
0.0009 
5. Batdispersed species 

Pipe ma 













F[3,87] = 47.97 
<0.0001 
6. Largearborealanimaldispersed species 

Thev ah 













F[2,88] = 4.50 
0.0368 
Quas am 















Byrs cr 













F[2,88] = 5.18 
0.0252 
Swar s2 













F[3,87] = 6.48 
0.0024 
Inga ve 













F[2,88] = 4.76 
0.0318 
Spon mo 















7. Agoutidispersed species 

Gust su 













F[4,86] = 17.73 
<0.0001 

Notes: Only the variables that were significant (P < 0.05) following backward elimination are presented. For remnant vegetation classes, both the negative exponential (Ex) and linear functions (Lin) were used in the model. All reported values in the 10 lefthand columns are R^{2} statistics resulting from regression against a single explanatory variable of distance from the various remnant vegetation classes or abiotic factors. Positive exponent: increasing density with distance from source; negative exponent: decreasing density with distance from source. Source includes: distance from the forest (Forest), distance from the nearest largeleaved monocot (Monocot), distance from the nearest remnant tree larger than 5 m in height (Tree), and distance from the nearest shrub or sapling 2.5–5 m in height (Shrub). Habitat: see note^{††}. Biotic = partial R^{2} for all biotic variables. Abiotic = partial R^{2} for all abiotic variables. Total = total R^{2} for the (multiple) nonlinear regression model. F = F statistic; n = 90.
^{†} Species: Bana gu = Banara guianensis, Byrs cr = Byrsonima crassifolia, Cecr in = Cecropia insignis, Coch vi = Cochlospermum vitifolium, Cord al = Cordia alliodora, Gust su = Gustavia superba, Hyba pr = Hybanthus prunifolius, Inga ve = Inga vera, Lueh se = Luehea seemannii, Mico ar = Miconia argentea, Pipe ma = Piper marginatum, Psyc gr = Psychotria grandis, Quas am = Quassia amara, Spon mo = Spondias mombin, Swar s2 = Swartzia simplex var. ochnacea, Thev ah = Thevetia ahouai, Trem mi = Trema micrantha, Tric ga = Trichospermum galeotti.
^{‡} Square root of distance from the forest.
^{§} Distance from the nearest largeleaved monocot (Musa sp., Heliconia sp.).
^{} Square root of distance from the nearest isolated oldfield tree larger than 5 m in height.
^{¶} Ln(distance from the nearest shrub or sapling between 2.5 m and 5 m in height).
^{††} Abiotic variables include: Live = biomass of live Saccharum (g), Dead = biomass of Saccharum litter (g), Fire = time since last wildfire (months). All abiotic variables were ln(x + 1) transformed, except fire; time since the last wildfire was not transformed.