I am constructing a survey with around 10 multi-item scales. My piloting indicates that one of the most important multi-item scales (COP) has very poor reliability (.49), but there is an item in another scale (REX) which seems to hang together semantically with this COP scale. Including this item in the problem scale brings the reliability coefficient up to .74, but taking it away from the REX scale brings that alpha down to .53...
Because I don't really want to go back to the drawing board and rewrite the items in the COP scale, and then re-pilot it, this seems to offer me a shortcut.
I'm wondering if computing a variable (COP) with this item that will also be included in another variable (REX) is possible, or if it is just not done that way. Can single items be used multiple times - that is, in multiple scales of a single survey?
Any opinions or help much appreciated.
