Development of a Self Report Stress Scale Using Item Response Theory-I: Item Selection, Formation of Factor Structure and Examination of Its Psychometric Properties

Abstract

Introduction: Due to the absence of Turkish psychometric devices assessing stress, in the present study it was aimed to develop a stress scale, and examine its basic psychometric properties. Methods: Current study included two processes, formation of item pool and examination of psychometric properties of the selected items through three studies. In the first study, 611 individuals aged between 18 and 77 responded to 130 selected items. In the second study, 2223 individuals aged between 18 and 68 responded to 80 items. In the third study, 1969 individuals aged between 18 and 79 responded to the final form of 36 items. Further, in study 3 for criterion related validity 163 individuals completed the Coopersmith Self-Esteem Inventory, 113 individuals completed the Beck Anxiety Inventory, 104 individuals completed the Hospital Anxiety and Depression Scale, 107 individuals completed the Beck Depression Inventory, and 265 individuals completed the Perceived Stress Scale. Moreover, in the investigation of test-retest reliability, 119 individuals took the final form of the test after 2 weeks, and 111 individuals took the final form of the test after 3 weeks. Results: In the first study, out of 130 items, 54 that showed item-total score correlations below 0.30 were excluded from the scale. Fiftyseven items were preserved exactly, and 19 items’ sentence structures were changed. Furthermore, by adding 4 new items, 80 were prepared for the second study. In the second study, two factors structure namely “Physiological Reactions/Strain” and “Psychological/Cognitive Appraisals” sub-dimensions were identified, and 36 items were selected via Item Response Theory representing these sub-dimensions. In study 3, exploratory factor analysis provided strong support for our hypothesized two factors structure. Confirmatory factor analysis indicated hypothesized model had a better fit to the data. Internal consistency coefficients were 0.94 for the entire scale, 0.90 for Physiological Reactions/Strain sub-dimension, and 0.91 for Psychological/Cognitive Appraisals sub-dimension. Correlation coefficients between the entire scale and other criterion scales ranged from 0.22 to 0.63. Test-retest correlation coefficients between the first administration of the scale, and the administrations at two and three week intervals were 0.88. Conclusion: Results showed that the scale has basic psychometric requirements provided that the scale will be supported by validity studies.