To demonstrate the explainability of the scales with fewer dimensions instead of the number of existing dimensions by ensuring that the scale structures created by explanatory factor analysis (EFA) are verified with confirmatory factor analysis (CFA). Data from the Nutritional Behavior Scale in Children, answered by the parents of 204 children with autism spectrum disorder (ASD) were used. EFA was performed with the data obtained from the scale. In the next step, the explained variance percentages and dimensions were determined and the model goodness of fit indexes were calculated with CFA. The dimensions with the lowest explained variance percentages were removed from the model, respectively, and three different scale models were tested. The variance explanation percentage of the first eight-dimensional model created with EFA was calculated as 72.68%. The food fussiness sub-dimension was removed and CFA was applied to the model again and new indices were calculated. Finally, the emotional under-eating sub-dimension was excluded from the model, resulting in a six-dimensional Child Eating Behavior Scale (CEBS). Goodness-of-fit indices of the CFA model established with six dimensions were χ2 / df; 1.545, AIC; 715,433 and RMSEA; 0.052 was found. It has been shown that the eight, seven, and six-dimensional scale models constructed according to the percentages of variance explained for CEBS are sufficient to explain the sample and that the six-dimensional scale model can be used for CEBS. Our study is the first to use competing models in confirmatory factor analysis in reducing scale dimensions.
Key words: CFA, EFA, competitive models, child eating behavior scale
|
