Stress is defined as a person's interaction with their environment that is thought to threaten or affect an individual's potential, resources, and well‐being. The most popular instrument to assess perceived stress is the Perceived Stress Scale (PSS). Therefore, making a systematic review of studies testing the internal structure of PSS and conducting a Meta‐Analytic Confirmatory Factor Analysis (MACFA) on the database created with the information obtained from these studies are the aims of this research. A total of 76 samples from 57 unique studies were included in this database using various inclusion criteria (total N for PSS‐14 = 28,632, for PSS‐10 = 46,053). The correlated two‐factor model for PSS was confirmed by MACFA performed on the pooled correlation matrix generated by the random effects meta‐analysis. Findings of dimensionality analyses, factor loadings, omega values, and measurement invariance showed that the model that best explained the factor structure of PSS was the correlated two‐factor model. The strict measurement invariance of the PSS was achieved across age and clinical status, and the internal consistency was high according to the omega values. Several recommendations moving forward are discussed.
One of the most popular instruments used to assess perceived social support is the Multidimensional Scale of Perceived Social Support (MSPSS). Although the original structure of the MSPSS was defined to include three specific factors (significant others, friends and family), studies in the literature propose different factor solutions. In this study, we addressed the controversial factor structure of the MSPSS using a meta-analytic confirmatory factor analysis approach. For this purpose, we utilized studies in the literature that examined and reported the internal structure of the MSPSS. However, we used summary data from 59 samples of 54 studies (total N = 27,905) after excluding studies that did not meet the inclusion criteria. We tested five different models discussed in the literature and found that the fit indices of the correlated 3-factor model and the bifactor model were quite good. Therefore, we also examined both models’ factor loadings and omega coefficients. Since there was no sharp difference between the two models and the theoretical structure of the scale was represented by the correlated three factors, we decided that the correlated three-factor model was more appropriate for the internal structure of the MSPSS. We then examined the measurement invariance for this model according to language and sample type (clinical and nonclinical) and found that metric invariance was achieved. As a result, we found that the three-factor structure of the MSPSS was supported in this study.
15 April 2016 in Hacettepe University Journal of Education
In this study, whether item position effects lead to DIF in the condition where different test booklets are used was investigated. To do this the methods of Lord’s chi-square and Raju’s unsigned area with the 3PL model under with and without item purification were used. When the performance of the methods was compared, it was revealed that generally, the method of Lord’s chi-square identified more items with DIF than did the method of Raju’s unsigned area. The differentiation of the booklets with respect to item position resulted in a higher number of items displaying DIF with item purification conditions. Based on the findings of the present study, to avoid the occurrence of DIF due to item position effects, it is recommended to position the same items across different booklets in similar locations when forming different booklets.
The aim of this study is to examine the factor structure of Turkish version of the Multidimensional Scale of Perceived Social Support (MSPSS) and to analyze its psychometric properties through the bifactor-ESEM framework. Using a convenience sample of 1124 Turkish adults, seven different models were tested. The results supported the superiority of a bifactor-ESEM solution that included three specific factors (family, friends, and significant others) and a general factor. In addition, bifactor indices showed that the general factor of MSPSS is not strong enough and its multidimensional structure is supported. For the bifactor-ESEM model, strict measurement invariance was achieved according to the gender variable. Our results supported convergent validity for the general and specific factors of the MSPSS, which were found to be associated with measures of distress, loneliness, and resilience. As a result, MSPSS is a valid and reliable measurement tool with its bifactor-ESEM model.