Wording Effect In GHQ-12

According to the World Mental Health Survey of the World Health Organization,the lifetime prevalence of global mental disorders is between 18.1% and 36.1%.People with mental disorders have a considerable proportion in all regions and age groups.Because of the difficulty of quantifying individual mental health and the existence of clinical misjudgment,it is not easy to find so many patients with mental disorders.The GHQ-12 scale is a type of General Health Questionnaire.It maintains reliability and validity while also being simple and time-saving.It is translated into multiple languages and is widely used in different countries to identify and screen mental disorders.At the same time,the dimensions included in the GHQ-12 scale have been controversial.From the perspective of psychological traits,there are multiple assumptions about the dimensions of psychological traits,such as single-dimensional and threedimensional.From the perspective of wording effects,the measurement results of the scale are greatly affected by positive and negative wording.How to exclude the effect of wording effects and find out the true psychological traits reflected by the scale from the data has also become a difficult problem that needs to be discussed.Because the Likert scale is used in the psychological measurement performed in this article,and ordered data is obtained,multiple correlation coefficient methods are used in the research on the correlation between topics.Before performing factor analysis on the data,this article uses parallel analysis to determine the number of factors between 2~4.Through EFA analysis,it is found that the correlation between topics is greatly affected by the wording effect.If only 2 are extracted,the factor reflects two different wordings: positive and negative;when comparing the 3-factor model with Graetz’s model,it can be found that it is different from the previous model,only because the wording was changed in question 11.In this way,the 3-factor model is still greatly affected by the wording effect.In the part of confirmatory analysis,this paper establishes a general CFA model,several different CTCM,CTCU,ESEM models.The general CFA model has low validity and cannot meet the normal standards of the fitting indicators,but the other kinds of models perform well.Compared with CTCM and CTCU models,ESEM models can actively identify the small cross-loading of the problem on different factors,and rotate like EFA models if needed,so their performances of fitting indicators are even better,but the complexity are higher.After all,this paper finds that the CTCM,CTCU,and ESEM models all have good fitting results.Among them,the CTCM model with single psychological characteristics and two wording effects also considers the simplicity of the structure.In addition,comparing the negative wording factors of this model with the ESEM model,we can also find that the two psychological traits of “anxiety” and “loss of confidence” have opposite loads on the negative wording factors.The relationship of negative wording needs further study.Finally,from the perspective of measurement equivalence,the GHQ-12 scale has only approximately the same dimensions in the data of primary and secondary school students in different regions,genders,and grades,and does not have measurement equivalence.Although GHQ-12 have been confirmed on many documents that it can be used as a measurement tool in different groups,its measurement results for different populations are not completely equivalent.