Stats help!

[DCJD24]

---

Members don't see this ad.

Forgive me for my lack of knowledge - I am an undergraduate. I was running regular linear regressions on my thesis data but my advisor told me that it wouldn't give me the unique contributions of each predictor.

Any ideas on what analysis I would need to use?

 

[cara susanna]

It's difficult to look at the unique contributions unless you run a single design analysis for each variable, which would increase alpha inflation. You could look at part and partial correlations when you run the regression, which tells you if the regression is being muddied by suppressor or inhibitor effects.

Edit: Or what Ollie said. 😉

 

[Ollie123]

You will likely need to provide more information on this for us to help. If you ran them separately he/she may have been suggesting you run a single model. Conversely, if you were running them as a single model he/she may have been suggesting you run them separately. I'm guessing the former since they may "overlap" in terms of what they predict, hence the statement that you don't know if they are contributing anything unique, but its nothing more than a guess at this point.

I'd just ask them how they suggest you model it. They likely had something in particular in mind and that is probably easier than trying to get us to translate🙂

 

[Jegg]

Multiple linear regression? Read this site.

I doubt they would have you using non-linear regression models, but you can read up on those also...

 

[Jegg]

Here is a paper on multivariate, multiple nonlinear regression if you are interested (or if your variable relationships do not meet the criteria needed for a linear model)...

 

[aequitasveritas]

Run a multiple and examine beta weights