中文版长处和困难问卷的信度概化分析

发布时间：2018-10-21 16:21

【摘要】：长处和困难问卷(Strengths and Difficulties Questionnaire,SDQ)是评估青少年儿童情绪和行为问题的常用评估工具。自1997年发表以来在世界范围内广泛应用,至今已有超过4000项实证研究使用了该工具。很多研究已经表明,长处和困难问卷具备良好的信效度(R.Goodman,1997;R.Goodman et al.,1998;R.Goodman,Renfrew,Mullick,2000)。然而,该问卷在国内使用过程中,品行维度和同伴维度得分的信度在不同的研究中差异较大(品行维度的α系数范围在0.35~0.848之间,而同伴维度的α系数则在0.001~0.602之间)。究竟是什么因素造成如此大的差异?本文拟使用信度概化技术对该问卷的信度系数这种不一致的表现进行定量分析,探讨既往研究之间信度系数的差异,并检验样本特征和研究特征对信度系数的影响。本研究对46项研究中的三种信度:α系数、重测信度系数、分半信度系数分别进行异质性检验,对α系数进行调节效应分析。概化分析结果表明,随机效应模型估计的信度均值接近0.7或大于0.7。三种信度及各因子的异质性显著,品行因子表现较其他因子更不稳定(Q=5486.716),同伴因子的异质性程度相对其余因子较低(Q=1641.000),表明同伴因子的信度表现出“较稳定的低”。混合效应模型的调节效应分析结果表明:年龄标准差、总样本量、均分、文献来源、出版物类型、样本类型都会对总量表α有预测作用;出版年份和文献来源对品行问题分量表得分的α系数有预测作用;均分和抽样地区对同伴关系分量表得分的α有预测作用。该工具在跨样本的使用上信度不稳定。结论:在未来应用长处和困难问卷时,应注意这些样本特征和研究特征对信度的影响。
[Abstract]:The strengths and difficulties questionnaire (Strengths and Difficulties Questionnaire,SDQ) is a common tool for assessing adolescent children's emotional and behavioral problems. It has been widely used in the world since its publication in 1997. So far, more than 4000 empirical studies have used the tool. Many studies have shown that the strengths and difficulties questionnaire has a good reliability and validity (R. Goodman / 1997 / R. Goodman et al.,1998;R.Goodman,Renfrew,Mullick,2000). However, in the process of using the questionnaire in China, the reliability of the scores of conduct dimension and peer dimension was significantly different in different studies (the 伪 coefficient of the conduct dimension ranged from 0.35 to 0.848, and the 伪 coefficient of the peer dimension was between 0.001 and 0.602). What makes such a big difference? This paper intends to use the reliability generalization technique to quantitatively analyze the inconsistent performance of the reliability coefficient of the questionnaire, to explore the differences of reliability coefficients among previous studies, and to examine the influence of sample characteristics and research features on reliability coefficients. In this study, three kinds of reliability of 46 studies, 伪 coefficient, retest reliability coefficient and split-half reliability coefficient, were tested for heterogeneity, and the adjustment effect of 伪 coefficient was analyzed. The results of generalizability analysis show that the estimated reliability of stochastic effect model is close to 0.7 or greater than 0.7. The heterogeneity of three kinds of reliability and each factor was significant, the character factor was more unstable than other factors (QN 5486.716), and the heterogeneity of peer factor was lower than that of other factors (QN 1641.000), which indicated that the reliability of peer factor was "stable low". The results of adjustment effect analysis of mixed effect model show that age standard deviation, total sample size, mean score, literature source, publication type and sample type can all predict the total amount of table 伪; The year of publication and the sources of literature can predict the 伪 coefficient of the score of the sub-scale of conduct problems, and the mean score and the sampling area can predict the score of the sub-scale of peer relationship. The reliability of the tool in cross-sample use is unstable. Conclusion: in the future application of the strengths and difficulties questionnaire, we should pay attention to the influence of these sample characteristics and research features on reliability.
【学位授予单位】：广州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：B848

【参考文献】