I agree with Vania. The technique just depends on your response/dependent variable. If you have categorical responses for you dependent variables, just use a categorical response model like multinomial logistic regression...with the appropriate tests.

Also, never worry about multicollinearity a priori. All it does is (potentially) lead to an increase in your standard errors because you will be using less independent variation/information if multicollinearity is present. If your variables of interest are all statistically significant, then there is nothing to worry about. If your control variables are collinear you have nothing to worry about.

Use regression models with spatial diagnostics using GeoDa for example.

Regression analysis would be helpful. You can use several other statistical analysis for interpretation like PCA analysis and such diagrammatic representation can be done using software like Primer etc.

I would suggest using Multiple Correspondence Analysis or related methods. In contrast to regression analyses it can handle categorical data and you have no problems with multicollinearity and the like. You can do it with R, for example. More info here:

http://factominer.free.fr/

I agree with Vania. The technique just depends on your response/dependent variable. If you have categorical responses for you dependent variables, just use a categorical response model like multinomial logistic regression...with the appropriate tests.

Also, never worry about multicollinearity a priori. All it does is (potentially) lead to an increase in your standard errors because you will be using less independent variation/information if multicollinearity is present. If your variables of interest are all statistically significant, then there is nothing to worry about. If your control variables are collinear you have nothing to worry about.

Using regression model, but it depends on data type such as ordinal, interval, or scale.