ok question - I am doing factor analysis and PCA on a dataset with 1000 columns, and there's roughly n=230 shared between any given column.
I'm not intending this to be 'robust' - would it make sense to squint at the factors and remove Qs that don't seem to belong?
Conversation
Replying to
You don’t need to delete questions. The Exploratory factor analysis should show you which items could be clustered under certain factors. At the end, some items may not fit any of the factors and it is fine.
1
6
Replying to
Perform a regression to determine if there is relevance to the question you are asking.
Replying to
In biology, outliers are often where something surprising and interesting is happening. Could be the same here.
Replying to
PCA already finds which factors aren’t relevant by assigning them to the less significant eigenvalues
5
Replying to
I’m not sure I understand the question, although data is how I earn money:))
1
Replying to
Squinting can’t be measured, you’re better off putting on glasses intended to blur your vision at a certain factor. Then you can at least quantify the removal of Q’s by working in the blur factor of the glasses you chose