Colin,
Here is a possibility to deal with this statistically (sorry not an spss solution but an r open source solution)
Code your drivers as x1 thru to x???? (use the no of drivers you are modelling here as a set of x1 to x? variables, for example driver 1 gets x1, driver 2 gets x2 and so on) and code your dependent variable as y,
delete all the peripheral information in your dataset and of course save as a newly named file in your c drive
move your y variable next to the ID (second in the spss data package) and x1 to x8 next to the y variable
Install r and download just two packages for use in this situation.
Download
1 Rcommander
2.Rlaimpo
load Rcommander first
(it supports spss data importing into r) and all you are using this package for is to create a dataset for rlaimpo the second package to use. This will bring up a new window and this is like an old fashioned user interface but it does the job (importing files into R is hideous!!!!!!)
by default, r commander will import spss datasets using the name "Dataset" and create them as "objects" which can be analysed by other packages
load rlaimpo and then use the following code in the main window
calc.relimp(Dataset, type = "lmg", diff = FALSE, rank = TRUE, rela = TRUE)
this will partition out multi-collinearity and use all the variables you have coded as x's and regress them against y. the algorithm uses every single combo and order of x variables (eg x1, then x1 with x2, then x1 with x2, x3 etc) regressed on y and averages them over each of these to give a "decomposed" outcome, essentially removing a lot of the multic-collinearity influence you get with spss ols regressions
HTH paul
then load
> Valdiserri, Colin <
[hidden email]> wrote:
>
I am creating a rather simple model using a linear multiple regression and have one
standardized beta that is negative.
I am rather confident it is due to the multicollinearity, but the model has to move forward
as it is per the clients request for tracking measures. How should I treat this negative
beta when using it in simulations for the client report? One colleageue suggested zeroing it
out, but that will not fly with the client. Can I just use th absolute value?
Any thoughts are greatly appreciated.
Colin
_______________________________________________________________________
The information transmitted is intended only for the person or entity to which it is
addressed and may contain confidential and/or privileged material. Any review,
retransmission, dissemination or other use of, or taking of any action in reliance upon,
this information by persons or entities other than the intended recipient is prohibited. If
you received this in error, please contact the sender and delete the material from any
computer.
_______________________________________________________________________
Locked
