When trying to understand and analyze huge datasets that contain categorical and numeric variables, what methods can I use to cluster, or more generally, view in a meaningful way to find correlation.
For example, if my dependent variable is numeric and I have 4 categorical and 4 numeric independent variables, how can I discover which independent variables most highly impact the dependent variable.
I would appreciate even just suggestions of modeling methods to look into.
The two modeling methods I'd look at first are decision trees and regression with ANOVA. I'd fit regression models with and without different variables and compare them using an ANOVA table to see if the extra variables added predicting value to the model. Decision trees are handy, as they put the more significant variables higher in the tree and lower values lower in the tree.
There are obviously more complex solutions, but these two are fairly easy to pull off and will give you a lot of information very quickly.
Also try to look at logistic regression analysis, since it makes no assumptions about the distribution of input variables. Linear regression has an assumption that the residuals need to be normally distributed, but not logistic regression.