I'm building a multi-variable regression model (I have data for each variable, by month, from 1/1/87 through 8/1/05). Here's my problem: some numbers are simply counts, while others are more descriptive in nature (percentages). Since I will be running a regression where some of the independent variables are large whole numbers (such as 450,000) while others are percentages, do I have to treat the independent numbers in some fashion prior to running the regression software? Meaning: do I have to normalize the independent variables? I recall in grad school I would sometimes use the natural log (ln) of the independent variables rather than the numbers themselves, but I can't recall why.