Development of a “Perceived Stress Scale" Based on Classical Test Theory and Graded Response Model

Abstract

The main purpose of this study is to develop a perceived stress scale based on Classical Test Theory (CTT) and Graded Response Model (GRM); to compare the parameters of the items in the scale that are tried to be developed according to both models, and to determine under which theory the measurement tool produces more reliable and valid results according to these compared item parameters. The item discrimination parameter value calculated according to CTT ranges from 0.472 to 0.735. On the other hand, item discrimination parameter values calculated under GRM vary between 1.062 (Item 15) and 2.606 (Item 2). Correlations between item thresholds were tested and the calculated correlation coefficients were; r =0.840 for β1 (p<0.01), r = 0.947 for β2 (p<0.01), r = 0.713 for β3 (p<0.05), and r = 0.559 for β4 (p<0.05) respectively. It can be assumed that these values not only support the item invariance of the items in the scale, but also show that the GRM is suitable for the data used for the scaling of the items. The reliability coefficient of the scale, in terms of internal consistency, was calculated as 0.919 according to the CTT, while the marginal reliability coefficient calculated as 0.931 according to GRM. Both reliability coefficients are quite high. In conclusion, there is a high correlation between the item parameters calculated according to both approaches, and the perceived stress scale (PSS) that is being developed can measure the desired features.The main purpose of this study is to develop a perceived stress scale based on Classical Test Theory (CTT) and Graded Response Model (GRM); to compare the parameters of the items in the scale that are tried to be developed according to both models, and to determine under which theory the measurement tool produces more reliable and valid results according to these compared item parameters. The item discrimination parameter value calculated according to CTT ranges from 0.472 to 0.735. On the other hand, item discrimination parameter values calculated under GRM vary between 1.062 (Item 15) and 2.606 (Item 2). Correlations between item thresholds were tested and the calculated correlation coefficients were; r =0.840 for β1 (p<0.01), r = 0.947 for β2 (p<0.01), r = 0.713 for β3 (p<0.05), and r = 0.559 for β4 (p<0.05) respectively. It can be assumed that these values not only support the item invariance of the items in the scale, but also show that the GRM is suitable for the data used for the scaling of the items. The reliability coefficient of the scale, in terms of internal consistency, was calculated as 0.919 according to the CTT, while the marginal reliability coefficient calculated as 0.931 according to GRM. Both reliability coefficients are quite high. In conclusion, there is a high correlation between the item parameters calculated according to both approaches, and the perceived stress scale (PSS) that is being developed can measure the desired features.