. 2021 Jul 25;196(4):1153–1166. doi: 10.1007/s00442-021-04992-x

Table 2.

Statistical models explaining variance in soil C measured among 141 grassland transects

Hypothesis	Model #	Model description	Model formula with P values^a	R²	AIC	df model, df error
1	1	Full model	Soil C–species richness^* + grassland type^* + clay^ns + prairie dog presence/absence^ns	0.41	337.3	5, 135
1	2	Constituent model 1	Soil C–species richness^***	0.21	374.3	1, 139
1	3	Constituent model 2	Soil C–grassland type^***	0.33	351.7	2, 138
1	4	Constituent model 3	Soil C–clay^*	0.03	403.8	1, 139
1	5	Constituent model 4	Soil C–prairie dog presence/absence^***	0.14	386.9	1, 139
1	6	Best model	Soil C–species richness^* + grassland type^* + species richness X grassland type^**	0.44	330.0	5, 135
2	7	Models for species subsets	Soil C–native species richness^***	0.17	381.2	1, 139
2	8	“	Soil C–exotic species richness^**	0.04	402.1	“
2	9	“	Soil C–native perennial graminoid species richness^***	0.28	361.5	“
2	10	“	Soil C–exotic perennial graminoid species richness^***	0.09	394.9	“
2	11	“	Soil C–native perennial forb species richness^***	0.11	391.4	“
2	12	“	Soil C–exotic perennial forb species richness^**	0.06	398.9	“
2	13	“	Soil C–native annual forb species richness^ns	0.0	408.7	“
2	14	“	Soil C–exotic annual forb species richness^ns	0.0	409.2	“
2	15	“	Soil C–Andropogon gerardii 1/0^***	0.27	364.6	“

See Figs. 1 and 2 for graphical depictions of relationships and more statistical information. See ESM2 for the data file for use in re-running these models to investigate the coefficients and effect sizes

The best model is in bold

^a***P < 0.001; ^**P < 0.01, ^*P < 0.05, ns not significant