Re: statistical question (somewhat obscure)
Richard Bell (rcbell@rubens.its.unimelb.edu.au)
Wed, 24 Jun 1998 09:13:37 +1000
At 22:57 23/06/98 +1000, you wrote:
>Richard,
>
>Thanks for the reply. Regarding your post:
>
>>Bob
>>
>>The answer to question is 1. providing element & construct loadings have
>>been scaled in the same metric (always hard to prove). However you may want
>>to break it up by also plotting the means (which will have almost certainly
>>been conditioned out in the principal compoenents) as an orthogonal axis,
>>which would be strictly analagous to plotting testitems by discrimination
>>and difficulty
>
>
>I'm lost. Do you mean that plotting the element and construct means along a
>line/axis? What would this axis be orthogonal to, the element-construct
axis?
>
>Regards,
>
>Bob
>
>
Bob
One axis, say the horizontal, one represents the first factor loadings. The
other, vertical axis, is perpendicular to this passing through zero on the
horizontal axis. Both elements and constructs will be represented as points
in this two-dimensional space.
Richard
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%