# Another question re: correlation and regression

#### arikel

##### New Member
I seem to be getting confused between correlation and regression a bit...

I'm trying to identify the relationship between a series of X variables and a series of Y variables. If an X variable is correlated with the Y variable, is it then appropriate to do a regression to "get more from the data", or is this only the case if my hypothesis are about predictive relationships?

Soo confused!

#### Dr.D

##### New Member
If you are seeking to see whether x-variable predicts or influences a y-variable, regression is appropriate. If you are just looking at relationships, then correlation is the way to go.

Regression and correlation are mathematically the same. Regression identify relationships too, however, you can also predict scores on Y-variables based on knowing scores on x-variables (Y = a + bx).

Do you know that a linear regression produces the same information as a a simple correlation? Yup it does, so dont worry.

#### arikel

##### New Member
I guess I could try to predict scores, but the theoretical backing isn't too great so I thought correlaton would be good as its so exploratory...

Perhaps it would be better to do a stepwise regression in this case? That way I get the correlation coefficients and exploratory regression analyses too.

What do you reckon?
The other problem is that some of the X predictor variables are correlated with one another... am I right in thinking that multicollinearity only occurs when it's a significant relationship of like .90, as opposed to .3 etc? Thanks again for your help #### Dr.D

##### New Member
You dont have to do stepwise regression.

Are you looking to see whether the x-variables affect the y-variables, you can do the multiple regression. However, multicollinearity is possible if any two x-variables are correlated with each other at .8 and above (not .3). If intercorrelations are less than .8, you should be good to do a multiple regression: A set of x-variables is related to each y-variable.