I am finishing an instrument that will subtype individuals based on constellations of various etiological risk factors and want to include only the factors (subscales) most relevant to the overall model.
Hello. Try regression. Each risk factor can be defined as criterion and your subscales can be predictors. In this way you can evaluate the importance of the predictors for your criterion and you can select the most relevants
Dear Lia,
The selection process of the subscales of a psychometric Battery must be theory-driven. It is very important to clarify which is your nomological net and the theoretical framework. After that, it is crucial to obtain empirical evidences of diverse sources of validity, basically related to content, criterion (convergent/discriminant, or/and decision) and factor structure. I agree with Cristian Opariuc-Dan about to use regression procedures, but sometimes an "automatic" or only-empriical guided selection strategy for predictors falls into capitalization on chance problems. Perhaps you could reach a very good prediction with some variables, but this model may be a tailored one for your sample. This risk could be controled by systematic replication, trying to confirm the best predictors. In my opinion, you have to combine empirical results obtained from validity evidences and the condiitions of your theoretical framework.
All the best.
- Badges
- Science topic
- Similar topics
- Psychology
- Psychometrics