Factor Analysis

#1
I ran the factor analysis, but I want to make sure I have the groupings right. I am attaching my analysis and the rotated component matrix has purple groupings that I created. What I am puzzled about is what to do with the last "ImpOrientation" because it is all by itself.

Also, am I supposed to choose only one group per factor (column)? So, for the first factor, I would only choose the ones that are on top? If items are below .5, do I throw them out? Is the first column grouping on top (ImpTreatManage to ImpPE one group?).

Also, is it as simple as looking at these groupings or am I supposed to look also at all the other tables?

Obviously, it's my first time running a factor analysis.

Thanks
 
#2
I would suggest trying to extract either 5 or 6 factors (under extraction button). The scree plot suggests 5 factors (there is an elbow at 6, "cut" above that elbow). ImpOrientation may then end up being associated with one of the other factors. You may also try another extraction method such as principal axis factoring (pull-down in the same tab).

Interpreting the loadings depends on the goal of your research. Often factor analysis is used for the purposes of studying the factors (aka "latent variables") and the observed variables are a way to indirectly measure them. If that's the case, then approach naming the factors from the perspective of "what might be an unobservable variable (e.g. attitude towards ____) that might explain why these 6 variable are closely associated with each other?" Since the first 6 items are highly associated with that factor, focus on those when naming the factor.

The cutoff for "meaningful" loadings is subjective, but .3 or .4 are typical values. It's more important for interpretation that variables clearly load one one factor and not others. This is part of a concept called simple structure. For example ImpRespAssigned does load on two factors, but clearly is more correlated with F3. ImpLearning and ImpEvidence are similar. ImpPatInterview is more troublesome. However, this pattern matrix generally exhibits good simple structure. I suspect that if you try fewer factors this might improve more.

The pattern matrix is what you want to look at. Those coefficients are correlations of variables with the common factors that you attempt to name.