act对马赛克的测量不变性：有和没有act测试适应的学生社会情绪学习评估

Social Emotional Learning Assessment forStudents With and Without ACT TestingAccommodations Yao Sun, Jill McVey, and Cristina Anguiano-Carrasco Accurate and equitable assessments are essential for making valid comparisons andunderstanding the impact of social and emotional learning interventions across diverse studentpopulations, including students with disabilities (Klingner & Edwards, 2006). Measurement Mosaic™ by ACT®: Social Emotional Learning Assessment (hereafter referred to as Mosaic) adopts a five-factor model, measuring skills related to Sustaining Effort, Getting Along withOthers, Maintaining Composure, Keeping an Open Mind, and Social Connection, each of whichaligns one-to-one with the Big Five personality framework (Walton et al., 2023). Our studytested measurement invariance for a subset of Mosaic items between two groups: students withand without testing accommodations on the ACT, which are given in line with the AmericansWith Disabilities Act. Students receiving accommodations for disabilities include students withneurodevelopmental disabilities (e.g., learning disabilities, ASD), physical-sensory disabilities Method Participants The sample consisted of high school students in grades 9 through 12 who completed theMosaic assessment between August 2020 and January 2021. Participants with and withouttesting accommodations were matched based on demographic characteristics including gender,race/ethnicity, and grade level when Mosaic was taken. The matched dataset contains datafrom 477 students without testing accommodations and 477 students with testing American, 15 as bi/multiracial, 6 as Asian, 14 as American Indian or Alaska Native, and 2 asNative Hawaiian or Other Pacific Islander. An additional 16 students did not report their race orethnicity. The sample included 262 students in 9th grade, 139 in 10th grade, 57 in 11th grade, The sample of students with testing accommodations had a similar demographic composition.Among them, 207 identified as female, 264 as male, and 2 as another gender; 4 students didnot report their gender. Regarding race and ethnicity, 336 students identified as White, 50 asHispanic or Latino/a, 33 as Black or African American, 15 as bi/multiracial, 2 as Asian, 13 asAmerican Indian or Alaska Native, and 1 as Native Hawaiian or Other Pacific Islander; 27 Materials Mosaic is a multi-method online assessment designed to measure five core social andemotional (SE) skills: Sustaining Effort, Getting Along with Others, Maintaining Composure,Keeping an Open Mind, and Social Connection. These skills are aligned with the Big Fivepersonality traits and are assessed using three item formats: Likert-type items, situationaljudgment tests (SJTs), and forced choice (FC) items. Likert items ask students to rate howmuch they agree with specific statements, while SJTs present brief scenarios followed bypossible responses that students rate based on how likely they are to act that way. In contrast,FC items present sets of equally positive statements and ask students to choose which Data Analysis Guided by measurement invariance theory (Vandenberg & Lance, 2000), we adoptedconfirmatory factor analysis (CFA) to determine whether SE assessment items yield equivalentresults for students with and without ACT testing accommodations. We conducted a series ofincreasingly robust measurement invariance analyses in R using the Iavaan package (Rosseel,2012). For each of the five factors measured using Mosaic, we tested three levels ofmeasurement invariance: configural, metric, and scalar (Meredith, 1993; Vandenberg & Lance,2000). Configural invariance refers to the condition in which the same factor structure is present invariance is considered the minimum requirement, while scalar invariance is a more stringentcriterion for interpreting scores across groups in the same way (Putnick & Bornstein, 2016). To begin with, we ran CFA for each factor for all Likert and SJT items and retained only theitems that yielded good factor loadings (> .40) for further analyses. The measurement model foreach factor yielded satisfactory model fit (CFIs and TLIs > .90, RMSEAs < .08, SRMRs < .06). Then we tested the configural invariance of each of the five factors to confirm that the factorstructure of each one was consistent across groups (students with and without testingaccommodations). For each factor, we specified the two groups in the CFA model, then used achi-square test to compare the model fits before and after specification of the two groups. The Next we tested the metric invariance of each factor, meaning we tested whether factor loadingswere equivalent across groups. Similarly, for all factors, we constrained factor loadings so theywere equal across groups and ran a series of chi-square tests to compare the model fits of the Finally we tested scalar invariance by constraining the intercepts of the items so they wereequal for both groups. The results of the chi-square tests showed that the mode