*2.3. Statistical Analyses*

The important value (*IV*) of *i*-th species in each treatment group in 2008 and 2014 were calculated using the following equation:

> *IV*(*i*)= (relative density(*i*)+relative frequency(*i*)+relative prominence(*i*)) / 3

The change in *IVs* for each species was directly calculated using the following equation:

$$\text{The changed value}\_{(i)} = IV\_{(i)(2014)} - IV\_{(i)(2008)}$$

Two-way analysis of variance (ANOVA) with season and treatment as fixed factors was applied to examine the effects on leaf and soil N, P and the N:P ratio based on averages. Community N:P stoichiometry (N:P stoichiometry means N, P, and the N:P ratio, community N:P stoichiometry means N:P stoichiometry of the plant community) weighted by *IV*(*i*) (2014) was also examined in the same manner (hereafter, community N:P stoichiometry is the leaf N:P stoichiometry weighted by the *IV*(*i*) (2014)). Meanwhile, the marginal means of the main factors were calculated and recorded. Three-way ANOVA with season, treatment, and different groups (broadleaves vs. conifers, trees vs. shrubs, ectomycorrhizal species (EM) vs. arbuscular mycorrhizal species (AM), broadleaves including broadleaved trees and shrubs; the data for conifers were not included in tree and EM groups) as fixed factors were used to analyze effects on leaf N, P, and the N:P ratio. Nonsignificant interactions were removed from the analysis. Leaf P, the leaf N:P ratio and soil N:P stoichiometry in our data were apparently of right-skewed distribution, and they were naturally logarithmically transformed before performing ANOVA; untransformed data are shown in the figures.

The data were separated according to (1) different seasons and (2) different seasons and different classified groups. Least squares discrimination (LSD) one-way ANOVA was applied to assess differences between treatment groups, and *IV*(*i*) (2014) was calculated as weightiness. Differences between N:P stoichiometry averages in our study and previous studies were evaluated using the one-sample *t*-test. The Kolmogorov–Smirnov and Levene's tests were employed to assess the normality of the residuals and the homogeneity of variance, respectively. All figures were created using Origin 9.0, and all data were analyzed using SPSS 20 (SPSS, Inc., Chicago, IL, USA). Significance was determined at the 0.05 level.
