Table 2.

Results of long‐term trend estimates using linear mixed‐effects model. Models were developed with the lme package (Pinheiro et al., 2015) with species as factor with random slope and intercept. The ‘original data’ used the uncorrected growth data for the nine species, the ‘corrected data’ use growth data adjusted for the difference of the shuffled trends from zero (see Text S1), the third model corrects for the nonuniform age distribution by adding age as second explanatory variable, and the last model excluded the three species (Brachystegia cynometroides, Brachystegia eurycoma and Chukrasia tabularis) with clearly clustered age distributions. Note that all models excluded the three species (Melia azedarach, Sweetia fruticosa and Afzelia xylocarpa) that have negative biases due to mortality effects (see Groenendijk et al., 2015). See Text S1 for details and exact model formulation, and Table S2 for the full outcome of various models, Values in black are significant at P < 0.05

	Canopy trees			Understory trees			Number of species
	Trends (% per decade)	P‐level	AIC	Trends (% per decade)	P‐level	AIC
1. Original data	0.8%	0.491	5194	0.8%	0.356	4489	9
2. Corrected data	2.1%	0.023	5193	1.3%	0.046	4492	9
3. Adding age as explanatory variable	1.6%	0.040	5164	1.0%	0.105	4430	9
4. Excluding species with age bias	2.3%	0.008	3373	1.3%	0.036	3022	6

Values in bold are significant at P < 0.05.