High values for independent variables

TheGreyP82

New Member
I have a dependent variable with rather small values (0 to 50) and several independent variables which can take on very high values (up to billions).
Because of this I have a heavily clustered correlation scatter plot which makes it difficult to spot any direction of the regression line. However, there is a correlation of about 0.5. When running the regression I get extremely small values for the beta coefficients which makes interpretation difficult. What would be a good way to run a regression with this data so that I can get some more plausible coefficients?

TheGreyP82

New Member
Haven't thought about that, but how do I interpret such a regression coefficient if I divide assets by 1 million?
For example, the number of pages in a financial report is my dependent variable, and the total sum of assets my independent variable.
Would it be something like: a 1/1000000 increase in assets increase the number of pages by 0.285? instead of a \$1 increase increase the number of pages by 0.000000285?

Buckeye

Member
Yes, you just change the interpretation based on a unit change in the independent variable. For instance, if you left it as 1,200,000,000 you would be interpreting the mean change in y for a dollar increase in x. Versus the mean change in y for every billion dollar increase in x. There are a several ways to go about it. The idea is to just change the scale.

hlsmith

Not a robit
I agree with Buckeye and I would add when creating scatter plots you can usually reformat axes to add more space and also use transparency in the plots to better see areas of more or less density.