Can an exploratory factor analys from an original psychometric testing of an instrument be used in a confirmatory factor analysis in a new language, i.e. when translating, adapting and testing the instrument for validation and reliability in another language?
Exploratory factor analyses are data driven and the results will invariably reflect some idiosyncrasies of the "sample" of persons whose data were analysed. When moving from one language to a next language you will likely want to test a strong theoretically derived model of how the instrument works, rather than an empirically derived model that is likely to change from one situation to the next.
This does not mean that EFA has no place in instrument validation across languages/cultures. One can expect results from an exploratory factor analysis that closely matches the "theoretical structure" to have a reasonable chance of being replicated in a new language (given the assumption that the construct measured by the instrument has a similar meaning in the new group). In addition, EFA can provide insight into reasons why a CFA model does not fit well (e.g. identify manifest variables that have loadings on more than one factor). If the results of the EFA makes good theoretical sense then it might be worthwhile to include it in a subsequent CFA.
I hope this helps.
Exploratory factor analyses are data driven and the results will invariably reflect some idiosyncrasies of the "sample" of persons whose data were analysed. When moving from one language to a next language you will likely want to test a strong theoretically derived model of how the instrument works, rather than an empirically derived model that is likely to change from one situation to the next.
This does not mean that EFA has no place in instrument validation across languages/cultures. One can expect results from an exploratory factor analysis that closely matches the "theoretical structure" to have a reasonable chance of being replicated in a new language (given the assumption that the construct measured by the instrument has a similar meaning in the new group). In addition, EFA can provide insight into reasons why a CFA model does not fit well (e.g. identify manifest variables that have loadings on more than one factor). If the results of the EFA makes good theoretical sense then it might be worthwhile to include it in a subsequent CFA.
I hope this helps.
Technically, yes it can.
Of course you have to be cautious of sample-dependency in the EFA, as Gideon outlined. The size of the sample-dependency relies heavily on the sample size. If your EFA-sample size was considerable, the resulting Factor Model would be a very good start for CFA on a new dataset (from a different population); especially when you can back the results from the EFA up with theoretical arguments.
- Badges
- Science topic
- Similar topics
- Psychology
- Psychometrics