2014, Vol. 22, No. 10, 1660 1668 Advances in Psychological Science DOI: 10.3724/SP.J.1042.2014.01660 研究方法 (Research Method) * 基于不对称区间估计的有调节的中介模型检验 1 2 2 2 ( 1, 510320) ( 2 /, 510631) 有调节的中介模型是中介过程受到调节变量影响的模型 评介了基于 Bootstrap 不对称置信区间和贝叶斯不对称可靠区间进行有调节的中介模型检验的 3 种方法, 包括亚组分析法 差异分析法和系数乘积法 模拟研究发现, 偏差校正的百分位 Bootstrap 置信区间和无先验信息的贝叶斯可靠区间在有调节的中介模型检验中表现相当, 都优于百分位 Bootstrap 置信区间的表现 建议使用系数乘积法进行第一阶段或第二阶段被调节的中介模型检验, 使用差异分析法进行两阶段被调节的中介模型检验, 并用一个实际例子演示如何用不对称区间估计检验有调节的中介模型 随后评述了 3 种有调节的中介模型检验方法在国内心理学的应用现状, 并展望了检验的拓展方向 有调节的中介 ; Bootstrap; 贝叶斯 ; 差异分析法 ; 系数乘积法 B841 1 前言 (mediation) (moderation),, (,,, 2012),, 3,,,, (moderated mediation) (mediated moderation), : 2013-09-16 * (14YJC1900 03) (GD13 CXL01) :, E-mail: fangj@gdufe.edu.cn ( (2013)) (X M Y) Z, Z, 3 ( 1) (X M) (M Y) (X M M Y) (Edwards & Lambert, 2007) ( Bootstrap ),,,, (Preacher, Rucker, & Hayes, 2007; Edwards & Lambert, 2007; Hayes, 2013a, 2013b; Wang & Preacher, in press;, 2012;,, 2014a, 2014b), Bootstrap 3, 1660
10 : 1661 1 ( Hayes, 2013b),,,, MPLUS WinBUGS ; 3, 2 基于 Bootstrap 的有调节的中介模型检验 1(a), (1) (2) M i1 a1x a2z a3xz (1) M Y i2 c 1X c 2Z c 3XZ bm 1 (2) Y (1)(2) M, (3) Y ( i2 bi 1 1 a2bz 1 c 2Z) [( c 1 c 3Z) ( a1 a3z) b1] X b 1 M (3) Y (a 1 +a 3 Z)b 1 c 1 c 3 Z Bootstrap (Preacher et al., 2007; Edwards & Lambert, 2007; Hayes, 2013a, 2013b; Wang & Preacher, in press;, 2012;,, 2014a), (), (), () 0, () 2.1 亚组分析法 Preacher (2007) (a 1 +a 3 Z)b 1 (conditional mediation effect), n, n, Z (Z h ) (Z l ), (a 1 +a 3 Z h )b 1 (a 1 +a 3 Z l )b 1 0, Preacher (2007),,,,, (a 1 +a 3 Z h )b 1 (a 1 +a 3 Z l )b 1 Hayes (2013a),, (a 1 +a 3 Z h )b 1 (a 1 +a 3 Z l )b 1,, ( ),,, (Edwards & Lambert, 2007);,, (,, 2014a;,,, Luthans, 2012;,,, 2012), Z (conditional value),, Z h Z l Z (Hayes, 2013a, 2013b),
1662 22, (,, 2014a) 2.2 差异分析法 Edwards Lambert (2007),, (a 1 +a 3 Z h )b 1 (a 1 +a 3 Z l )b 1 (a 1 +a 3 Z h )b 1 (a 1 +a 3 Z l )b 1 0,, Z, (mean sd), (Hayes, 2013a, 2013b), (Edwards & Lambert, 2007) (,, 2012;,, 2012) ( ) (, 2012)(2014a),, Z,,, Z 2.3 系数乘积法 Hayes (2013a, 2013b) Z,, ( a a Z ) b ( a a Z ) b a b ( Z Z ) (4) 1 3 h 1 1 3 l 1 3 1 h l a 3 b 1 (Z h Z l ) 0, Bootstrap a 3 b 1 (Z h Z l ), Z h Z l Bootstrap, a 3 b 1 (Z h Z l ) 0, a 3 b 1 Bootstrap, a 3 b 1 0, 任意 a 3 b 1 (Z h Z l ) ( 4) 0, Hayes (2013a, 2013b), a 3 b 1 Bootstrap, Hayes (2013a, 2013b) PROCESS (http://www.afhayes. com/), SPSS, ( c 1 c 3 Z h ) ( c 1 c 3Zl) c 3( Zh Zl), c 3 0, c 3 ( Zh Zl) 0, (Hayes, 2013a, 2013b), ( 1(b)), (b 1 +b 2 Z)a 1, a 1 b 2 0, (Hayes, 2013a, 2013b), ( 1(c)), (a 1 +a 3 Z)(b 1 +b 2 Z), ( a1 a3zh)( b1 b2zh) ( a1 a3zl)( b1 b2zl) 2 2 ( ab ab)( Z Z) ab( Z Z ) (5) 1 2 3 1 h l 3 2 h l Z (Hayes, 2013a, 2013b), Z ( 1 0), a 1 b 2 +a 3 b 1 +a 3 b 2 0, (Hayes, 2013b) 2.4 小结, Bootstrap,,,,,,, 3 基于贝叶斯的有调节的中介模型检验 Wang Preacher (in press) (Bayesian method) ( 1(a)), (1) (6)~(8) M i ax (6) M M im i1 a2z (7) a a1 a3z (8), (2) (9)~(12)
10 : 1663 Y i cx bm (9) Y iy i2 c 2Z (10) cc 1 c 3Z (11) b b 1 (12) Wang Preacher, Z (i M i Y ) (a b c'), ( ab=(a 1 +a 3 Z)b 1 ) ( c 1 c 3 Z ),, ( 1(c)), (12) b=b 1 +b 2 Z WinBUGS MPLUS (Markov chain Monte Carlo, MCMC),, MCMC (Gibbs sampling) ( Gibbs MCMC, (, 2011;,,, 2011;,,, 2014)),, Wang Preacher (in press),, MPLUS ( XZ MZ), WinBUGS, (6)-(12) ( 2),, MPLUS WinBUGS, MPLUS Gibbs ( 10000~20000), WinBUGS Wang Preacher (in press),, MPLUS ANALYSIS POINT=MEAN ( 2), Wang Preacher (in press), Wang Preacher Y,, (WinBUGS MPLUS 2), Wang Preacher ( ),, Bootstrap, ( ), 4 基于 Bootstrap 和贝叶斯的检验比较 Bootstrap, Bootstrap (Percentile Bootstrap), (Bias-Corrected) Bootstrap, Bootstrap Bootstrap (, 2011) Wang Preacher (in press),, Bootstrap Bootstrap, Ⅰ,, Bootstrap, Bootstrap Wang Preacher (), Wang Preacher, Bootstrap,,,,,,,
1664 22,, Bootstrap 5 示例, (dysfunction team behavior) (X)(negative affective tone) (M) (team performance) (Y), (nonverbal negative expressivity) (Z),, 2 () Cole (Cole, Walter, & Bruch, 2008; Hayes, 2013b) 60, 277 X M Z, XZ MZ MPLUS6.0 2 Cole (2008) Bootstrap (MPLUS 1) a 1 b 2 95%[ 0.89, 0.03], 0, (b 1 +b 2 Z)a 1 = 0.27 0.3Z Z 0.55 (mean sd) 0.01(mean) 0.54(mean+sd), 0.110.27 0.43 Bootstrap c 3 [ 0.01, 0.83], 0, (WinBUGS MPLUS 2) Bootstrap,, (X) (Y), (X)(M) (Y), (Z) (X) (Y), 6 应用现状分析 Bootstrap, 2014 6 9, 10, ( ), 12 Bootstrap (,,, 2009;,, 2011;,, 2012;,, 2012;,,, 2012;,, 璟,, 2012;, 玥,, 2013;,,,, 2013;,,,,, 2014; 惿,, 2014;, 2014;,, 2014), 1, 12, Bootstrap 2007 (Preacher et al., 2007; Edwards & Lambert, 2007), (2009), 7 (58%), 10 (83%), 2013, Bootstrap (),, (5 ) (3 )(2 ), (3 ),,
10 : 1665 表 1 Bootstrap 有调节的中介应用研究的文献分析 2012 mean±sd 2012 mean±sd 2013 mean±sd 2014 mean±sd 惿 2014 mean±sd 2009 mean±sd 2011 mean±sd 2012 mean±sd 2012 mean±sd 2014 mean±sd 2013 mean±sd 2014 mean±sd ( 2 ), Hayes (2013a, 2013b), (4 ), 3, mean±sd, ( 2.2 ) (2014a), (a 1 +a 3 Z)(b 1 +b 2 Z), 1 (),, a 1 b 2 +a 3 b 1 +a 3 b 2 0, ( 2.3 ), 7,, c 3 0, ( 2.3 ),,, ( (b 1 +b 2 Z)a 1 = 0.27 0.3Z),, Z,,, c 1 c 3 Z 7 讨论与拓展 Bootstrap 3,,,,,,,,,,, Z, W, (a 1 +a 3 Z) (b 1 +b 2 W)(Preacher et al., 2007;, 2012; Wang & Preacher, in press) Hayes (2013b) Bootstrap (2012) Bootstrap,,, 1
1666 22 ( )( ),,, (,, ),,, (,,, 2012), Mackinnon Fairchild (2008, 2009), XM XMZ, X M Z,, Mackinnon Fairchild, Ⅰ (Mackinnon, 2008; Fairchild & Mackinnon, 2009),,,, (,, 2014a), (, 2012;, 2012;,, 2014a),, ;,,, 方杰, 张敏强, 李晓鹏. (2011). 中介效应的三类区间估计方法. 心理科学进展, 19, 765 774. 胡惿, 顾雪英. (2014). 使命取向对职业满意度的影响 - 职业弹性的中介作用及工作资源的调节作用. 心理科学, 37, 405 411. 胡阳, 范翠英, 张凤娟, 谢笑春, 郝恩河. (2014). 青少年网络受欺负与抑郁 : 压力感与网络社会支持的作用. 心理发展与教育, 30, 177 184. 李锐, 凌文辁, 柳士顺. (2009). 上司不当督导对下属建言行为的影响及其作用机制. 心理学报, 41, 1189 1202. 李锡钦. (2011). 结构方程模型 : 贝叶斯方法 ( 蔡敬衡, 潘俊豪, 周影辉译 ). 北京 : 高等教育出版社. 李燕萍, 涂乙冬. (2011). 与领导关系好就能获得职业成功吗? 一项调节的中介效应研究. 心理学报, 43, 941 952. 梁建. (2014). 道德领导与员工建言 : 一个调节 - 中介模型的构建与检验. 心理学报, 46, 252 264. 刘东, 张震, 汪默. (2012). 被调节的中介和被中介的调节 : 理论构建与模型检验 ( 高中华译 ). 见陈晓萍, 徐淑英, 樊景立 ( 编 ). 组织与管理研究的实证方法 ( 第二版, pp. 553 587). 北京 : 北京大学出版社. 刘乐平, 高磊, 杨娜. (2014). MCMC 方法的发展与现代贝叶斯的复兴 纪念贝叶斯定理发现 250 周年. 统计与信息论坛, 29(2), 3 11. 牛更枫, 郝恩河, 孙晓军, 周宗奎. (2013). 负性生活事件对大学生抑郁的影响 : 应对方式的中介作用和性别的调节作用. 中国临床心理学杂志, 21, 1022 1025. 申传刚, 马红宇, 杨璟, 刘腾飞. (2012). 上司不当督导与下属绩效 : 反馈寻求行为和学习目标定向的作用. 心理学报, 44, 1677 1686. 隋杨, 陈云云, 王辉. (2012). 创新氛围 创新效能感与团队创新 : 团队领导的调节作用. 心理学报, 44, 237 248. 隋杨, 王辉, 岳旖旎, Luthans, F. (2012). 变革型领导对员工绩效和满意度的影响 : 心理资本的中介作用及程序公平的调节作用. 心理学报, 44, 1217 1230. 王永跃, 段锦云. (2014). 人力资源实践对员工创新行为的影响 : 心理契约破裂的中介作用及上下级沟通的调节作用. 心理科学, 37, 172 176. 王永跃, 朱玥, 王铜安. (2013). 心理契约破裂, 工作满意度与建言行为 : 神经质的调节作用. 心理科学, 36, 1459 1463. 王宇清, 龙立荣, 周浩. (2012). 消极情绪在程序和互动不公正感与员工偏离行为间的中介作用 : 传统性的调节机制. 心理学报, 44, 1663 1676. 温忠麟, 刘红云, 侯杰泰. (2012). 调节效应和中介效应分析. 北京 : 教育科学出版社. 温忠麟, 叶宝娟. (2014a). 有调节的中介模型检验方法 : 竞争还是替补? 心理学报, 46, 714 726. 温忠麟, 叶宝娟. (2014b). 中介效应分析 : 方法和模型发展. 心理科学进展, 22, 731 745. 叶宝娟, 温忠麟. (2013). 有中介的调节模型检验方法 : 甄别和整合. 心理学报, 45, 1050 1060. Cole, M. S., Walter, F., & Bruch, H. (2008). Affective mechanisms linking dysfunctional behavior to performance in work teams: A moderated mediation study. Journal of Applied Psychology, 93, 945 958. Edwards, J. R., & Lambert, L. S. (2007). Methods for integrating moderation and mediation: A general analytic framework using moderated path analysis. Psychological Methods, 12, 1 22. Fairchild, A. J., & MacKinnon, D. P. (2009). A general
10 : 1667 model for testing mediation and moderation effects. Prevention Science, 10, 87 99. Hayes, A. F. (2013a). A simple test of moderated mediation. Manuscript submitted for publication. Retrieved May 7, 2013, from http://www.afhayes.com/ Hayes, A. F. (2013b). An introduction to mediation, moderation, and conditional process analysis: A regression-based approach. New York: Guilford Press. MacKinnon, D. P. (2008). Introduction to statistical mediation analysis. Mahwah, NJ: Earlbaum. Preacher, K. J., Rucker, D. D., & Hayes, A. F. (2007). Addressing moderated mediation hypotheses: Theory, methods, and prescriptions. Multivariate Behavioral Research, 42, 185 227. Wang, L., & Preacher, K. J. (in press). Moderated mediation analysis using Bayesian methods. Structural Equation Modeling. 附录 1: 基于 Bootstrap 的有调节的中介模型检验的 Mplus 程序 DATA: FILE = teams.txt;! team.txt VARIABLE: NAMES =x m z y;! USEVARIABLES = x m z y xz mz;! CENTERING IS GRANDMEAN (x m z);! x m z DEFINE: xz = (x-0.035)*(z-(-0.01));! x z, xz mz=(m-0.047)*(z-(-0.01));! m z, mz ANALYSIS: BOOTSTRAP=5000;! bootstrap 5000 MODEL: m on x (a1);!m x y on x (c1) m (b1) z xz(c3) mz(b2);! y x m z xz mz MODEL CONSTRAINT: NEW (ind); ind= a1*b2;! OUTPUT: CINTERVAL (BCBOOTSTRAP);! bootstrap : bootstrap, OUTPUT cinterval (bcbootstrap) cinterval (bootstrap) 附录 2: 基于贝叶斯的有调节的中介模型检验 WinBUGS # model {for (i in 1:N) {m[i] ~ dnorm (mean.m[i], prec.m) mean.m[i] <- i1 + a1*x[i] #(6) y[i] ~ dnorm (mean.y[i], prec.y) mean.y[i] <- iy[i] + c.prime[i]*x[i]+ b[i]*m[i] #(9) iy[i] <- i2 + c2.prime*z[i] #(10) c.prime[i] <- c1.prime + c3.prime*z[i] #(11) b[i] <- b1 + b2*z[i] } #(12) #( ) i1 ~ dnorm (0, 1.0E-6) a1 ~ dnorm (0, 1.0E-6) i2 ~ dnorm (0, 1.0E-6)
1668 22 c2.prime ~ dnorm (0, 1.0E-6) c1.prime ~ dnorm (0, 1.0E-6) c3.prime ~ dnorm (0, 1.0E-6) b1 ~ dnorm (0, 1.0E-6) b2 ~ dnorm (0, 1.0E-6) prec.y ~ dgamma (0.001, 0.001) prec.m ~ dgamma (0.001, 0.001) # theta <- a1*b2 } Mplus 程序 : 1 Mplus ANALYSIS OUTPUT, ANALYSIS: estimator = bayes;! chain=2;! 2 Markov process = 2;! Markov point=mean;! fbiter = 30000;! Markov 30000 OUTPUT: tech1 tech8 ;!tech1 ; tech8, PLOT: type=plot2;! Markov :, MODEL PRIORS Moderated Mediation Model Analysis Based on Asymmetric Interval FANG Jie 1 ; ZHANG Minqiang 2 ; GU Honglei 2 ; LIANG Dongmei 2 ( 1 School of Humanities and Communication, Guangdong University of Finance & Economics, Guangzhou 510320, China) ( 2 Center for Studies of Psychological Application & School of Psychology, South China Normal University, Guangzhou 510631, China) Abstract: Integrating moderation and mediation models has been a popular topic of study in social science methodology. A moderated mediation model is a combination of both moderation and mediation models. When a mediation effect is moderated by a moderator, the effect is termed moderated mediation and the model is a moderated mediation model. There are three common types of moderated mediation models: first-stage moderated mediation, second-stage moderated mediation, and dual-stage moderated mediation. Three methods of testing moderated mediation models are based on the Bootstrap asymmetric confidence interval and the Bayesian asymmetric credible interval. These methods include a subgroup approach, difference approach, and products of coefficients approach. The products of coefficients approach is recommended to analyze first-stage or second-stage moderated mediation models. The difference approach is recommended to analyze dual-stage moderated mediation models. A simulation study found that the Bayesian credible interval without informative prior yielded higher power than the Bootstrap percentile confidence interval, and comparable power to the Bootstrap bias-corrected confidence interval. In the present study, we used an example to illustrate how to conduct the proposed procedure by using MPLUS and WinBUGS software. MPLUS and WinBUGS programs are provided in the appendix to facilitate the implementation of the Bootstrap method to analyze moderated mediation effects. The programs can be managed easily by empirical researchers. To explore the present situation and problems of using three moderated mediation analysis methods in psychological researches in China, this article reviewed 12 studies published in Chinese psychological journals, retrieved from the CNKI database, since 2009 to 2014. Key words: Moderated mediation; Bootstrap; Bayesian; difference approach; products of coefficients approach