Table 1.

Model results for the fossil mammalian diversity–productivity relationship

	Estimate	Std. error/std. dev.	z	P
Continents
NPP	1.109	0.373	2.97	0.003
Area	0.535	0.081	6.59	<0.001
Stage duration	0.237	0.084	2.83	0.005
Continent identity	0.062	0.249
Stage midpoint	0.001	0.026
Regions
NPP	0.912	0.318	2.87	0.004
Area	0.628	0.089	7.04	<0.001
Stage duration	0.251	0.083	3.03	0.002
Region identity	0.346	0.588
Stage midpoint	0.004	0.062

Models were GLMMs with Poisson error functions, fitted across stratigraphic stages in either the two continents or the three focal regions. The response variable was fossil mammalian diversity (γ diversity estimates rounded to integers). Variables fitted as fixed effects were NPP in grams of dry matter per square meter per year, log-transformed; area (minimum convex hull around all mammalian locations in each stage, in square kilometers log-transformed); and stage duration (My, log-transformed). Variables fitted as random effects were continent or region identity (random intercept) and stage midpoint (age in Mya, random slope). The random effects therefore accounted for temporal trends and spatial structure in the data (see SI Materials and Methods and Table S2). Model statistics were as follows: model across continents: n = 12 observations; residual df = 5; marginal R² (variance explained by the fixed effects) = 0.68; conditional R² (variance explained by entire model, i.e., both the fixed and random effects) = 0.89; Akaike Information Criterion (AIC) = 116.6; model across regions: n = 17; residual df = 10; marginal R² = 0.39; conditional R² = 0.82; AIC = 183.3. Std. dev., SD of the variance estimate for random effects; Std. error, SE of the slope estimate for fixed effects.