Page 24 - Regression Guideline for AMC

Normal distribu6on of data
• Another pre-­‐modeling factor to consider is the distribu6on of the dependent variable (sales price). It is especially important that this variable be normally distributed.
• A normal distribu6on, also called a bell-­‐shaped curve, has many cases around the mean or average with fewer and fewer cases as you get further away from the mean in either direc6on.
• Before running the regression model, we should examine the distribu6on of the sales prices of the MLS proper6es to be included. If the distribu6on is too skewed away from normal or there are small numbers of cases that lie far away from all other cases, the regression model might not predict well or produced biased regression weights that lead to poor es6mates.
24