## Multiple Regression help:

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

sdh23
Posts: 2
Joined: Fri Mar 09, 2012 6:31 pm

### Multiple Regression help:

Hi,
I'm needing help with multiple regression. I am very new to MR and have very little time in which to teach myself the essentials. I have access to SPSS 19. I am testing to see if the number of organized violent events (dependent variable) in a given location increases based on a series of independent variables - population increase (0 - no; 1 - yes); vegetation productivity (continuous); export ban (0-no ban; 1- ban); land degradation (0 - low degradation; 1 - high degradation); dominant livelihood (1- pastoral farming; 2 - agro-pastoral farming). How do I prepare my data and/or decide which variables I need to include in the MR? I have read so far that I need to test for multicolliniarity and hetroscedacity - how do I do this? And once I have done that what would be the next step(s)? You'd be a lifesaver for helping me out, my project had to be changed last minute (not my fault!) and this is what I've been given to do instead!
Stephen
Posts: 43
Joined: Sun Mar 11, 2012 2:26 pm

### Re: Multiple Regression help:

Hi, Stephen
First recode dominant livelihood into 0/1 instead of 1/2.
Then look at pairwise correlations among all the regressors. If any of them exceeds 0.8, then multicollinearity is an issue.
Then use Analyze - Regression - Linear and start by including all the variables into the model.
The fact that they are all binary is OK.
Heteroscedasticity detection requires a few tests that are outside of SPSS. Or they can be implemented in SPSS but it will take some time.
Better multicollinearity diagnostics is also available in SPSS, but requires some knowledge on how to interpret it correctly (I am talking about VIF - variance inflation factor).
Eugene
Statistical Consultant