Statistical indices from bifactor models
Many instruments are created with the primary purpose of scaling individuals on a single trait. However psychological traits are often complex and contain domain specific manifestations. As such, many instruments produce data that are consistent with both unidimensional and multidimensional structur...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Instituto Peruano de Orientación Psicológica – IPOPS
2017-06-01
|
| Series: | Interacciones: Revista de Avances en Psicología |
| Subjects: | |
| Online Access: | http://ojs.revistainteracciones.com/index.php/ojs/article/view/51/html |
| Summary: | Many instruments are created with the primary purpose of scaling individuals on a single trait. However psychological traits are often complex and contain domain specific manifestations. As such, many instruments produce data that are consistent with both unidimensional and multidimensional structures. Unfortunately, oftentimes, applied researchers make determinations about the final structure based solely on fit indices obtained from structural equation models. Given that fit indices generally favor the bifactor model over competing measurement models it is imperative that researchers make use of the available information the bifactor has to offer in order to compute informative indices including omega reliability coefficients, construct reliability, explained common variance, and percentage of uncontaminated correlations. Said indices provide unique information about the strength of both the general and specific factors in order to draw conclusions about dimensionality and overall scoring of scales (and subscales). Herein, we describe these indices and offer a new module which easily facilitates their computation |
|---|---|
| ISSN: | 2411-5940 2413-4465 |