Item Wording Effects Of Core Self-evaluation Scale

The simultaneous use of positively and negatively worded items in the Core Self-Evaluation Scale(CSES)can have an impact on the scale’s factor structure,as well as its reliability and validity.Most researchers started from a factor analysis perspective and introduced methodological factors to explain the effect of item wording effects on scale structure,and few studies further explored whether the effect of wording effects on scale structure is stable.Recently,the use of constrained Factor Mixture Analysis(FMA)has been proposed to screen subjects affected by expression effects in order to restore the original structure of the scale.However,the FMA has only been validated on the Rosenberg Self-Esteem Scale and the intertemporal stability of this method has not been verified.Therefore,in this study,a cross-sectional study and a longitudinal study will be conducted to test the existence of item wording effects for the Core Self-Evaluation Scale,and on this basis,the validity of the constrained Factorial Mixed Analysis to eliminate the representation effects will be verified.The present study first conducted an item wording effect analysis on the Core Self-Evaluation Scale(CSES),using Mplus8.3 to conduct a series of confirmatory factor analyses on 975 university students.Results indicated that a model with one core self-evaluation factor and two wording effect factors(CTCM-PN)had the best fit(χ~2/df=2.75,CFI=0.98,TLI=0.97,RMSEA=0.04,SRMR=0.02).Then,the constrained Factor Mixture Analysis(FMA)method was used to eliminate the wording effects in CSES.Using structural equation modeling in Mplus8.3,the method’s effectiveness was examined.Results showed that the constrained Factor Mixture Model(FMM)effectively identified participants(n=55)whose responses were inconsistent with the logical expectations implied by the item wording,with an Entropy index of 0.65,indicating good classification accuracy.The resulting clean sample showed obvious unidimensionality in parallel analysis,with twelve items loading onto one core self-evaluation factor(mean loading=0.55).Comparing the full sample and clean sample using single-factor,two-factor,and random intercept item factor analysis models,the clean sample’s unidimensionality model provided a better fit.The correlation between the two factors in the two-factor model increased to 0.87,indicating a higher shared variance of 75.69%,while the method factor loading decreased to 0.14,All of the above results demonstrate the effectiveness of the constrained Factor Mixing Analysis(FMA)method for eliminating expression effects.Study 2 tracked 501 university students with two data collections separated by a 6-month interval,and used Mplus8.3 for longitudinal invariance test.The study aimed to examine the cross-time invariance of the best-fitting model CTCM-PN identified in Study 1,and also to test the cross-time stability of the effectiveness of the constrained Factor Mixture Analysis(FMA)method.By testing four nested models with gradually increasing restrictions,the results showed that gradually tightening parameter restrictions did not significantly deteriorate the model fit,and the absolute value of the fitting index did not change more than 0.01,demonstrating the longitudinal invariance of the wording effects and FMA.The conclusions of this study are as follows:1.The Core Self-Evaluation Scale(CSES)was influenced by both positive and negative wording effects,and the CTCM-PN model containing one core self-evaluation factor and two representational factors fitted best.2.The wording effects was stable across time and satisfied the stability characteristic of response style.3.The constrained Factor Mixture Analysis(FMA)could effectively eliminate wording effects from the scale,and its validity was reliable and stable.