Fig. 5From: Performance of alternative spatial models in empirical Douglas-fir and simulated datasetsMedian and 90% central quantile of sampling distribution of metrics (root mean squared error of prediction of the expected phenotype -RMSE-; residual variance relative deviation -RVRD- ; correlation between predicted and true simulated spatial effects -Scor-; and spatial variance relative deviation -SVRD-) by fitted model (bidimensional separable first-order autoregressive -AR- and bidimensional spline regression -splines-), spatial range (short and long), and plot structure (design, single-tree plot -STP- and multiple-tree plot -MTP-)Back to article page