Hi
One of the best methods for variable selection and generation predictive model its MLR Stepwise that exist in SPSS software. I try to use this method for selection of 80 variables to make MLR model, but I have critical problem with the number of variable that select for generation of model. SPSS generate the best model with r>0.95 but with 8 selected variables.
In my perspective this number of variable or independent variable its not good for MLR model. So I wanna to help me to reduce the number of Selected variable but with r>0.9 .
Best regards
Mohammad
Hi,
To change the value of r must operate on the syntax.
I don't know which version of SPSS you are using. Therefore, I post a link to the tutorial basic syntax of SPSS and syntax manual for SPSS 20.0.
Good job
http://www.spss-tutorials.com/command-syntax-reference/
Hi Mohamad,
apart from the high number of predictive variable, a major problem with MLR is the independence among predictors. In other words, beside model fitting, predictors must be checked for multicollinearity (i.e. correlation among predictors).
Indeed, a MLR model with correlated predictors can indicate how well the entire bundle of predictors predicts the outcome variable, but it may not give valid results about any individual predictor, or about which predictors are redundant with respect to others.
Before tryng different MLR models with different combinations of predictors, one should check for multicollinearity. A possible approach is Principal Component Regression, in which principal components (orthogonal, thus mutually independent) of the potentially predictive variables (i.e. your 80 variables) are first extracted by PCA, and then used are independent predictors in MLR analysis.
See Jolliffe(J Royal Stat Soc 1982, Series C 31 (3): 300–303) for further details.
Dear Guido
At the first thank you, I perform PCA and obtain 8 component. In my perceptive I cant use this component for generation of Independent MLR model because these component are not our first variable!
Its true or not?
Hi,
entering 80 variables and let the SPSS choose for you is completely a wrong way to go.
Researcher use Stepwise when things got confusing regarding few variales. The main guide to choose the variable is the literature and the researcher experience in the field. then researcher leave to the statistical pogram a window for some help based on correlation and significance level.
Good luck
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling