home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math.stat
- Path: sparky!uunet!charon.amdahl.com!pacbell.com!decwrl!spool.mu.edu!sdd.hp.com!saimiri.primate.wisc.edu!zaphod.mps.ohio-state.edu!moe.ksu.ksu.edu!mimbres.cs.unm.edu!constellation!reepicheep.gcn.uoknor.edu!richman
- From: richman@reepicheep.gcn.uoknor.edu (Mike Richman)
- Subject: Re: Factor Analysis& Rotation
- Originator: usenet@midway.ecn.uoknor.edu
- Sender: usenet@constellation.ecn.uoknor.edu (Usenet Administrator)
- Message-ID: <1992Nov4.230542.19346@constellation.ecn.uoknor.edu>
- Date: Wed, 4 Nov 1992 23:05:42 GMT
- Reply-To: richman@reepicheep.gcn.uoknor.edu (Mike Richman)
- References: <1992Nov3.113547.3959@memstvx1.memst.edu>
- Organization: Engineering Computer Network, University of Oklahoma, Norman, OK, USA
- Lines: 38
-
- In article <1992Nov3.113547.3959@memstvx1.memst.edu>, rajagopa@memstvx1.memst.edu writes:
- |> Dear friends,
- |> Could some of you stat buffs let me know the concept of
- |> factor analysis and rotation in plain english. I do not quite get the hang
- |> of it. I thank your responses and will post the compiled version on the net.
- |> Thanks/Raj
- |>
-
- There are several ways of approaching the rotation issue. One is to
- view it historically from a "simple structure" standpoint. In this
- case, a decent text to read is Thurstone's "Multiple Factor Analysis".
- What he says is that simple structure rotation attempts to find a
- linear transformation ("rotation") which results in repositioning the
- initial (unrotated) factors in hyperspace such that the variable
- configuration is simplified as much as possible. He lists 5 rules to
- help the graphical rotators of the 1940's achieve this by using the
- concept of a hyperplane. In doing simple structure rotation, one
- typically isolates subsets of highly interrelated variables which
- covary similarly. There is some overlap with the concepts of cluster
- analysis. With the advent of computers, hand rotation has become
- nearly a lost art and most analysts depend upon some algebraic
- expression which attempts to fulfill some of Thursone's concepts (ie,
- Varimax).
-
- Another way of looking at factor rotation is from a quasi-confirmatory
- view. In such a situation, one might wish to look for a predetermined
- factor loading structure by finding a least-squares fit between a low
- dimensional subset of factors and some preconceived loadings. This
- type of rotation is known as target rotation. The most straightforward
- description of the concepts behind rotation is probably Rummel's (1970)
- text, Applied Factor Analysis, Northwestern U. Press. I also have a
- review article on rotation of principal components, which includes a
- Monte Carlo test of the accuracy of available algorithms, should that
- interest you.
- ---
- Mike Richman !CIMMS/The University of Oklahoma
- richman@reepicheep.gcn.uoknor.edu !100 East Boyd, Room 1110
- 405-325-3460 (office) 325-7614 (fax)!Norman, OK 73019-0628 USA
-