Explore chapters and articles related to this topic
The Use of Nonparametric Item Response Theory to Explore Data Quality
Published in Steven P. Reise, Dennis A. Revicki, Handbook of Item Response Theory Modeling, 2014
Rob R. Meijer, Jorge N. Tendeiro, Rob B. K. Wanders
Because the AISP focuses on the monotonicity assumption, several studies showed that this procedure performs worse in recovering the correct dimensionality structure than other methods (Mroch & Bolt, 2006; Smits, Timmerman, & Meijer, 2012; van Abswoude, van der Ark, & Sijtsma, 2004). In particular, those studies revealed that Mokken scale analysis does not function well in conditions in which the traits correlate (Mroch & Bolt, 2006; van Abswoude et al., 2004), or in which the items load on more than one trait (Smits et al., 2012; van Abswoude et al., 2004).
A matter of convergence: Classical and modern approaches to scale development
Published in Francis Guillemin, Alain Leplège, Serge Briançon, Elisabeth Spitz, Joël Coste, Perceived Health and Adaptation in Chronic Disease, 2017
Both classical and modern test theory share some common aspects in the ascertainment of indicators of psychometric quality. This is particularly the case for structural validity, the very basis of the validity for PROMs, where evidence is presented that a set of items can be summated to provide a total score. All other indicators of psychometric quality are based upon this foundation. Using simulated data as an example, it has been shown that much the same results can emerge from either approach. The more recent analytical strategies such as structural equation modeling also offer a more integrated approach, similar to that found in Rasch analysis and IRT in general, such that several indicators—local response dependence, unidimensionality, group invariance—can all be ascertained at the same time. Modern test theory has some additional benefits, notably the ability to check if polytomous items are working as intended, and the Rasch model provides an interval scale transformation when data satisfy its assumptions. This may be one way of improving the confidence in the various responsiveness indices, although carrying out such calculations on the metric. On the other hand, modern test theory has little to add to the tests of external validity, other than providing a confirmation that the item set can be summated and, again for the Rasch model, confirmation that this raw score is a sufficient statistic (all you need for the estimate of the persons level of the construct) (Fisher, 1934). Without needing the transformation to the interval scale, the nonparametric Mokken scale model will also confirm sufficiency (Mokken, 1971).
Dimensionality, Item Response Theory, Effect Size Attenuation, and Test Bias Analyses of the Self-Importance of Moral Identity Scale (SIMIS)
Published in Journal of Personality Assessment, 2022
Paul K. Lutz, Brian P. O’Connor, Dunigan Folk
The dimensionality of the 10 SIMIS items was investigated using multiple methods, including: (1) parallel analysis; (2) Velicer’s minimum average partial (MAP) test; (3) sequential chi-squared model tests (SMT); (4) the Hull method; and (5) the empirical Kaiser criterion. These methods were recommended and described by Auerswald and Moshagen (2019) on the basis of Monte Carlo simulation tests of dimensionality methods. Dimensionality was also assessed via fit coefficients (RMSR and GFI) for the possible N-factor solutions, and via Mokken scale analysis. Mokken scale analysis is a non-parametric, bottom-up method of assessing dimensionality and of identifying subscales in item pools. We used the automatic item selection procedure in the Mokken R package (Van der Ark, 2007) to determine if two scales consistent with the SIMIS Internalization and Symbolization model for the items emerged from the bottom-up analyses. Multiple methods for assessing dimensionality were used because there is no gold standard, and because converging evidence from multiple methods is more credible than the findings for any one method.
Increasing Intent to Vaccinate for Seasonal Influenza
Published in Journal of Community Health Nursing, 2020
According to the developers, the Knowledge Scale was well suited for examining group differences. The creators utilized test-retest reliability to correlate scores from an initial test with the same test given two weeks later. Test-retest reliability for the scale was r = 0.70 (p < .01, N = 104). Loevinger’s scalability coefficient of H = 0.45 (weak: 0.3 ≤ H < 0.4, average: 0.4 ≤ H < 0.5 strong: 0.5 ≤ H < 1.0) indicated the scale had strong reliability. In addition, the researchers were able to construct a Mokken scale, which provided strong support for reliability of the scale as the Mokken scale had much more stringent assumptions compared with classical test theory (Zingg & Siergrist, 2012). Two studies were completed with the aim of formulating relevant knowledge items related to an individual’s decision to vaccinate. The authors were able to replicate the findings of the first study, thereby supporting the reliability of the scale.
Establishing the cut-off score for aggression on the Brief Psychiatric Rating Scale-Excited Component (BPRS-EC) in schizophrenia patients
Published in Psychiatry and Clinical Psychopharmacology, 2019
Seon-Cheol Park, Eun Young Jang, Gyung-Mee Kim, Ajit Avasthi, Sandeep Grover, Andi Jayalangkara Tanra, Takahiro A. Kato, Kok Yoon Chee, Mian-Yoon Chong, Afzal Javed, Chay Hoon Tan, Norman Sartorius, Naotaka Shinfuku, Yong Chon Park
In the statistical analyses, the Mokken scale analysis [10] was used to measure the scalability of the BPRS-EC. In the Mokken analyses, the coefficient (H) of homogeneity (or scalability) indicated the contribution degree of each item to the overall measurement of aggression severity in schizophrenia patients. Based on the definition of Sijtsma and van der Ark [11], 0.30 ≤ H < 0.40 constitutes a weak scale, 0.40 ≤ H < 0.50, and H ≥ 0.50 a strong scale. In addition, the association between the BPRS-EC and BPRS-18 total scores was determined using the Pearson correlation. Statistical significance was set at P < 0.05 (two-tailed) in all tests. Furthermore, exploratory receiver operating characteristic (ROC) curve analyses were conducted to establish the optimal cut-off score for the presence of aggression on the BPRS-EC in schizophrenia patients. As described elsewhere [12], the ROC curve analyses were developed from the signal-detection theory and were frequently used in biological and behavioural studies [13]. To calculate overall predictor performance, the sensitivity and specificity of all possible threshold levels were considered to determine the optimal cut-off score, one that generated the lowest number of false positives and negatives. The Mokken scale analysis was conducted using R version 3.4.3 (https://www.r-project.org/) and the Pearson correlation and ROC curve analyses were conducted using SPSS version 24 (IBM Co., Armonk, NY, USA).