CAJ | 학술논문

以往对影响儿童同伴关系的因素的研究具有明显的元素主义倾向，无法对儿童同伴关系做出科学而又简约的预测。在系统总结以往研究的变量的基础上，筛选出了4大量表和6个可测量指标。统计结果显示，影响儿童同伴关系的因素依次是学校标签、能力特质和焦虑，三个因素对模型的方差累积贡献率为74．887％；验证性因素分析模型的拟合度高，路径系数均极其显著；三个隐变量能够解释ZLM62．2％的变异。结论表明儿童隐变量预测模型具有很好的构思效度，能较好地预测儿童同伴关系。
이왕대영향인동동반관계적인소적연구구유명현적원소주의경향，무법대인동동반관계주출과학이우간약적예측。재계통총결이왕연구적변량적기출상，사선출료4대량표화6개가측량지표。통계결과현시，영향인동동반관계적인소의차시학교표첨、능력특질화초필，삼개인소대모형적방차루적공헌솔위74．887％；험증성인소분석모형적의합도고，로경계수균겁기현저；삼개은변량능구해석ZLM62．2％적변이。결론표명인동은변량예측모형구유흔호적구사효도，능교호지예측인동동반관계。
At present, the domestic and alien researches on the factors affecting children＇s peer relationships show a general tendency of elementalism. These studies examined isolated relations between peer relationships and only a certain variable. Because there exist correlations between independent variables, a variable may be displaced by another variable with a greater effect size in the systemic factors when multivariate analyse is done. Therefore, this kind of research can not reach scientific conclusions. It is a systemic project to investigate the factors affecting on children＇s peer relationships. We selected 4 questionnaires and 4 measurable indices on the basis of previous researches, and adopted 478 students＇ available questionnaires and measurable indices from elementary schcol forth grade to sixth. There were no significant differences in terms of gender x^2（ 1, N = 478） =. 837, P =. 360） and whether or not class post x^2 （1, N = 478）= 1.414, P =. 234）. Kaiser-Meyer-Olkin （KMO） and Bartlett＇s Test of Exploratory Factor Analysis （EFA） showed that KMO score was . 767 and the value of Bartlett Test d Sphericity was 1171. 535 （P =. 000）, which indicated that the results of EFA were credible. EFA sorted 4 questionnaires and 4 measurable indices into three factors, i.e. latent variables. Total variance explained of EFA indicated that the first factor affecting peer relationships was school-label, the percentage of variance explained of which is 32. 765 %, and the sequent variabls were competence trait （23. 625 % ） and anxiety （18. 497 % ）. The above three factors＇ cumulative percentage of variance explained was 74. 887 % （please see Table 1 ）. Every rotated component loading coefficient was above . 729 （see Table 2）. Confirmatory factor analysis indicated that the model had a good model fit （see Table 3）. All regression weights and correlation between latent variables were significant. In all measurable variables, teacher acceptance, class post and SASC got the greatest factor loading correspondingly in school-label, competence trait and anxiety, as seen in Fig. 1. The factor loading reflects the dependence of measurable variables on latent variables or correlation between two kinds of variables. Multivariate linear the regression analyses were used to determine the relative contribution of three latent variables to ZML, the regression equation of which was significant （F（3,474）= 117.087, P =. 000）, and ZLL, regression equation of which was also significant （F（3,474） = 34. 411, P =. 0013）. Three latent variables could explain the variance of ZLM （Z score of Like Most ） by 62.2 %, and the variance of ZLL （Z score of Like Least） by 47.4%. The two models all achieved a good predictive effect, but the effect of latent variables predicting ZLM was better than ZLL. Standard regression coefficient indicated that school-label took on a most powerful predictive effect among three latent variables in two models. Furthermore, school-label and competence trait could predict ZML positively and ZLL negatively. On the country, anxiety could predict ZLL positively and ZML negatively, as seen in tables 4 and 5. In sum, these results indicate that three latent variables can explain a large proportion of variance of measurable variables, and account for the basic components affecting children＇ s peer relationships. It is feasible to use three latent variables to predict children＇ s peer rdationships scientifically and briefly.