Hi everyone,
I wanted to check the influence of independent variables (meteorological parameter) on my dependent variable (pollutant). But my independent variables are non-normal has multicollinearity and autocorrelation. Which regression analysis should i use in SPSS?
Can someone help me on this?
You can use Generalized Least Squares formalism in this case, and biased Model parameter estimation methods applicable under the scope of the GLS Regression framework.
Thank you for your suggestion. But i dont think SPSS provide GLS estimation. Can you suggest any other software?
Hello Nazneen,
Though it's not said in your query, I suspect that the autocorrelation issue is attributable to responses taken at successive times (minutes / hours / days...). OLS (ordinary least squares) regression would be a poor choice in that instance (as you no doubt recognize). The concerns about collinearity can be addressed in a variety of ways (examples: ignore if not excessive; drop redundant variables; combine variables via principal components or factor analysis into independent composites; use ridge regression); the "best" way depends on your specific research aims.
Linear mixed models would be one option; time series regression another; and, as Debopam Ghosh indicates, GLS can handle problems with correlated residuals.
FInally, all of these (and much, much more) are available in specific libraries in the freely available R system.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS