johorensis had outlying positive values for its difference in CON effect on growth P 2 –P 1 , for both the nonspatial and spatial models, but that was accounted for by the highly significant large effects in P 1 ( p ≤ .001) despite the corresponding effects in P 2 being very close to zero ( p > .05).
Adjusted R 2 ‐values were similar for the nonspatial and two spatial models of more interest, though CON coefficients were more often significant ( p ≤ .05) for spatial than nonspatial, and HET coefficients showed a slight trend in the opposite direction (Appendix S1 : Table S4). 3.3 Effect sizes dependence on species’ plot basal area abundance and density Regressing the 38 species’ CON effect sizes on growth in P 1 , whether at the individual species’ level they were significant or not, against plot BA (log 10 ‐transformed), showed that the nonspatial model with “ba + CON + HET” led to a substantially better fit ( p ≤ .001) than with “ba + ALL” ( p = .18) (Table 3a ) and accounted for the maximum adjusted and predicted R 2 of all nonspatial and spatial models.
Having “equal” instead of “larsm” canopy allocations led to also weak, marginally significant fits.