Hi List Members =
1. I would welcome members advice mainly on the correct interpretation of the below
data-reduction 6-component PCA correlation matrix, which was generated as a prelude to Multiple Regression Analysis.
2. To the PCA-inexperienced eye it would appear that as only three IVs (MDA:.709: FExMP: .652: TFs & Exhibs: .549) load solely - and reasonably heavily -on one component (the first), only these three should be carried forward as input to MRA. I might add that other research supports the importance of these 3 variables. Is my interpretation here correct and, if not, where does it go wrong ?
3. Finally, can members recommend any texts (with minimal maths) that provide advice on the interpretation of PCA output?. I am specially interested in guidance on how to choose between variables that load on a number of components. Intuitively, it would seem that, say,Variable X should be selected only if its loading on component 1 is much heavier than it is on 2 and 4. If there are in the field Rules for making such choices I would much like to know what they are and where I can access them.
With advance thanks =
Owen
INDEPENDENT VARIABLES: PCA CORRELATION MATRIX
Component
Variable 1 2 3 4 5 6
MDA .709 - - - - -
FExpS .701 - - - - -.360
FExpMP .652 - - - - -
NExpM .622 -.515 - - - -
NGeoAS .617 -.576 - - - -
NSChUsed .580 - - -.363 - .383
SMBMethods .570 - .302 - - -
MInfo S: TFs & Exhibs .549 - - - - -
- In/Out TM .546 - -.453 - - -
- VFRets .514 .312 -.324 - -.322 -
C/ PExp .526 .375 - - - -
CAt .525 - -.430 - .380 .307
SpStSk .516 - .393 -.422 - -
NQTU .492 - - .363 - -
APty .513 540 - - - -
C P & MBPg .374 - - .610 - -
Total FTF MA .455 - - -.522 .372 -
SUKJEu - .422 . 461 - .560 -
ContyOExp .425 420 - - - .470
[Extraction Method: Principal Component Analysis: 6 components extracted
|