Smid Sanne C, Hox Joop J, Heiervang Einar R, Stormark Kjell Morten, Hysing Mari, Bøe Tormod
Utrecht University, Utrecht, the Netherlands.
University of Oslo, Oslo, Norway.
Assessment. 2020 Dec;27(8):1901-1913. doi: 10.1177/1073191118803159. Epub 2018 Oct 5.
Emotional and behavioral problems among children and adolescents may be studied using the Strengths and Difficulties Questionnaire, containing five subscales, based on ratings by parents, teachers, or adolescents themselves. We investigate two measurement issues using data from a longitudinal sample of 8,806 participants aged 7 to 9 years and 11 to 13 years from the Bergen Child Study in Bergen, Norway. First, convergent validity of parent and teacher ratings is studied using a multitrait-multimethod approach. Second, longitudinal measurement equivalence is studied using confirmatory factor analysis, which requires us to deal with the considerable attrition. The multitrait-multimethod indicates not only good convergent validity but also considerable method variance for parents and teachers. The reliability and validity of some subscales are relatively low. Attrition analysis indicates that attrition is not missing completely at random, but estimation assuming missing at random makes no real difference. We conclude that assuming missing completely at random is acceptable. Comparing ratings by parents and teachers results in partial scalar equivalence. In addition, all subscales exhibit (partial) longitudinal scalar measurement equivalence. We recommend using latent variable modeling and not summated scales for longitudinal modeling using the Strengths and Difficulties Questionnaire.
儿童和青少年的情绪及行为问题可以通过优势与困难问卷进行研究，该问卷包含五个分量表，基于父母、教师或青少年自身的评分。我们利用来自挪威卑尔根儿童研究中8806名7至9岁以及11至13岁参与者的纵向样本数据，研究两个测量问题。首先，采用多特质多方法途径研究父母和教师评分的收敛效度。其次，使用验证性因素分析研究纵向测量等价性，这要求我们处理相当程度的样本流失问题。多特质多方法分析不仅表明了良好的收敛效度，还显示出父母和教师存在相当程度的方法变异。一些分量表的信度和效度相对较低。样本流失分析表明，样本流失并非完全随机缺失，但假设随机缺失进行估计并无实际差异。我们得出结论，假设完全随机缺失是可以接受的。比较父母和教师的评分会得出部分标量等价性。此外，所有分量表都呈现出（部分）纵向标量测量等价性。我们建议在使用优势与困难问卷进行纵向建模时，采用潜在变量建模而非求和量表。
