Factor analysis results - negative loadings and reverse coded items - Misc. Stats Topics Discussion

Misc. Stats Topics Discussion > Factor analysis results - negative loadings and reverse coded items

I ran an EFA in SPSS (oblique rotation) with 66 variables, all measured on a 7-point Likert scale. I specified 6 factors ahead of time (based on a parallel analysis) and suppressed coefficients < .40. My sample size is 240.

Two strange results:

1) My 66 items included 8 reverse-coded items, and all 8 of these loaded on one factor (and no other items did). Several of these 8 items appear to tap different constructs, so I suspect some methodological issue at play. But what? I double checked my reverse coding and it's correct. I also reran the EFA with the original (non-reversed items) and these 8 again loaded onto the same factor (suggesting my recoding was done correctly). Is this just a coincidence that all my reverse coded / worded items loaded onto the same factor?

2) One factor was comprised of 11 items, and all had negative loadings. Is that a suspicious finding? How do I interpret that?

August 22, 2013 |

Charles

STATS MAKE ME CRY

Misc. Stats Topics Discussion > Factor analysis results - negative loadings and reverse coded items

Post Your Reply