Stability is one of the most important characteristics in ecosystems. Based on the principle that the accumulative generation can enhance the linear correlation of two monotone incremental sequences and the methods of canonical correlation analysis, a grey canonical correlation model is developed for analyzing the effect of population on the stability of community, and the ratio of mn/mp （where mp is the total individual numbers of pests and mn their natural enemies ） used as a stability index of community. The modeling procedure consists of the following steps: (1) the individual number sequences of populations are rearranged according to increasing sequence of total individual numbers of pests in the community, and the sequences divided by the differences of various populations for dimensionless transform and accumulative generation; (2) the canonical variables are determined by the canonical correlation analysis of pests and their enemies, the pests and their natural enemies used as independent and dependent variables, and regression equations established based on the canonical variable pairs which meet the demand of linear fitting and inverse operation of accumulative generation; (3) the equations are linearly combined with a set of coefficients to ensure that the linear combination of canonical variable of pests in regression equations has the greatest correlation coefficients with the sequence of total individual numbers of pests in the model; and (4) a model of mutual conversion of total individual numbers between the pests and their natural enemies is developed by introducing the concept of conversion coefficient, which is called a grey canonical correlation model and can be used to analyze the effect of populations on the stability of community. The model has also been employed to analyze the stability of arthropod community in the tea plantation of Jinshan, Fuzhou, Fujian Province. The results is basically consistent with the data obtained from the field study, which well indicates that the model is feasible and applicable.