Investigating Dimensionality of Eskin's Attitudes Toward Suicide Scale with Mokken Scaling and Confirmatory Factor Analysis

Abstract

Attitudes toward suicide are often investigated by means of questionnaires, most of which are overly long or of low psychometric quality. Eskin's (2004) Attitudes Towards Suicide Scale is short and first investigations suggest good psychometric properties, but its factor structure has scarcely been explored. Hence, we examined this instrument by a data-analytic approach that combines Mokken scaling and confirmatory factor analysis (CFA). Results revealed 6 factors, of which 1 possessed only weak measurement properties. CFAs indicated only borderline fit of models found in prior research, but adequate fit for the scale structures revealed by Mokken scaling. Psychometric properties of these scales were satisfactory. Both the instrument as well as the methodological approach presented here can be recommended for further research.

Keywords:

• attitudes
• confirmatory factor analysis
• dimensionality
• item response theory
• Mokken scaling
• suicide

Notes

Note. Numbers in cells are item numbers, and numbers in parentheses are the respective scales' scalability coefficients H. minH _i  = minimum item scalability for an item to be selected to a scale. Sc = Mokken Scale.

Note. Factor names were taken from the six-factor solution of Eskin (2004). Item 16 was found to be unscalable and therefore excluded from analysis, but is still listed in the table. H = Loevinger's scalability coefficient for the complete scale; H _i  = Loevinger's scalability coefficient for a single item (i.e., item discrimination); α = Cronbach α; ρ = Molenaar-Sijtsma ρ.

Note. Split = Splitting the factor Suicide as a Solution from the factor Acceptability of Suicide (yes or no); χ² = χ² value of model fit; df = degrees of freedom; TLI = Tucker-Lewis index; CFI = comparative fit index; RMSEA = root-mean-square error of approximation; SRMR = standardized root-mean-square residual; numbers in brackets are 95% confidence intervals for the RMSEA.