Page 12 - Regression Guideline for AMC

Understanding regression mechanics The basic mul6variable regression equa6on is:
Yi =a=b1x1 +b2x2 +b3x3 +b4x4.....bnXn +ei
• In this equa6on Y is the actual value for some subject property i, the list of x’s are the predictors associated with that property’s characteris6cs, each b shows how much Y changes (i.e., value increases or decreases) for a 1 unit increase in X, and e is the amount of error associated with the predic6on of property Y. The a is a constant term that is usually not interpreted.
• The error term e in the equa6on is ouen called the “residual” because it reflects the residual (remaining) difference between the predicted and actual value of the subject property. A well-­‐fiRng regression model minimizes the residual term so the predicted values of the proper6es in the database are as close as possible to the actual values. Residual analysis is an important metric for evalua6ng model fit.
12