# Deleting variables from an exploratory factor analysis

I've carried out an exploratory factor analysis on a set of survey questions. The result is three factors, however the combination of variables in the third factor makes no sense theoretically. The first two factors do load on these variables (loadings are about .2). I'm going to use the factor scores as variables in further analyses.

Should I drop the variables in the third factor from the analysis, and redo the factor analysis with only two factors? Or keep the variables in the analysis, and just report (and use) two factors?

• How did you determine that the result is three factors? Did you do a parallel analysis? – Jeffrey Girard Apr 5 '17 at 0:03
• There was no parallel analysis, the number of factors was decided from subject matter knowledge - I should add though that these items weren't written with the intention of carrying out factor analysis. The first two factors loaded basically as we thought they would, but the third is an unexpected combination of variables. – edstatsuser Apr 9 '17 at 21:33
• Should I drop the variables in the third factor That depends. What is the purpose of your study? Are constructing a questionnaire with "simple" items (i.e. loaded by one factor each)? – ttnphns Apr 10 '17 at 14:57