Table 1.

A comparison of spatial regression methods based on the analysis of null expectation, by regressing allele frequencies evolving under a pure isolation-by-distance process against three explanatory variables (factors). N. models refers to the frequency (out of 20 simulations) with at least one significant (p < 0.05) regression slope, whereas N. coeff. shows the frequency (out of 60 coefficients) of significant coefficients. The Dist(H0) refers to the average Euclidian distances between the regression coefficient vector β and the null expectation (all slopes are zero).

	N. models	N. coeffs	Dist (H0)
OLS	0.65	0.38	0.140
TSA	0.75	0.33	0.170
PCNM	0.50	0.18	0.154
LagRES	0.40	0.13	0.076
LagPRED	0.40	0.20	0.113
SAR	0.70	0.35	0.118
CAR	0.70	0.38	0.138
MA	0.70	0.35	0.116