I am creating a psychometric instrument. I have asked several hundred respondents to rate 49 items (which colleagues and I came up with) on a 5-point LIkert scale. My aim was to reduce these to a small number of factors, similar to personality traits on personality questionnaires.

I carried out the analysis using SPSS and obtained 5 factors after orthogonal varimax rotation on which 25 items loaded (and I discarded the other 24 items). I used Laerd Statistics (statistics.laerd.com) which give instructions for PCA (principal components analysis), so I assumed I was doing this.

However, according to other sources (e.g. Haas-Vaughn, 2017, http://www2.sas.com/proceedings/sugi30/203-30.pdf and https://www.theanalysisfactor.com/the-fundamental-difference-between-principal-component-analysis-and-factor-analysis/) I may have done EFA (exploratory factor analysis).

Haas-Vaughn (p. 364) states 'If your goal is to estimate underlying factors and attach some meaning to those factors (as a form of construct validity, for example), then EFA is required.' I thought this was my goal.

So have I done EFA or PCA? Thanks.

Philip

More Philip A Fine's questions See All
Similar questions and discussions