Investigation of the pyschometric properties of academic ve general self concept inventory

Yandı, Alperen. Investigation of Psychometric Properties of Academic and General Self Inventory. Post Graduate Thesis, Bolu, 2013. In this study, psychometric properties of Academic and General Self Inventory (AGSI) consisting of six sub dimensions of Self-Description Questionnaire ? III (SDQ ? III) developed by Marsh (1987) to determine self-concept of the late adolescents whose ages are between 18 ? 26, was investigated. AGSI includes 62 items from the six sub dimensions. Following the steps of adaptation study, and the SDQ ? III was initially translated into the target language and then retranslated. Reviewing the translation, cohesion has been secured.After the translation process, AGSI was applied to 1945 participants from 26 universities in Turkey. Some of applications have been carried out by the researcher personally and some were conducted by using the inventory on via internet. After the application process, by using the ?expectation maximization? method in SPSS 13 missing data analysis was performed on data set. Having completed the missing data analysis, reliability coefficient of AGSI and sub-dimensions were calculated. Then, the exploratory factor analysis (EFA) was performed and the factor loadings of items were investigated. After the EFA, the confirmatory factor analysis (CFA) was performed on the data set by using LISREL 8.7. The fitting indexes of the model made up in accordance with the factors were checked. According to the first analysis, the fitting of the model was low. In the second stage of the analysis, the 14 items whose factor loadings were under 0,40 according to the EFA conducted in the initial step, were removed from the AGSI one by one and EFA was repeated. The best factor loadings regarding the items were gained by excluding 14 items in different dimensions from the data set. After this process, in the result of the repeated CFA, fit indexes were seen to be better than the first analysis and the generated model fit was identified to be high.