You are on page 1of 2

AI http://onlineed.appraisalinstitute.org/ai/OnlineEd/Onlineed.asp?Session_...

How the Model Arrives at Its


Results
Page 18 of 22

When you solved the size/price regression using a


graphical approach, the point was made that each
person would have a slightly different solution. When
looking at a scattergraph, there can be a number of
straight lines that appear to be equally reasonable in
terms of representing the average relationship
between the independent and dependent variables.
The regression model, however, needs a rule to “draw
the line;” otherwise, we could get various answers
based on the same set of data. And, it is important for
anyone using a regression model to know how it
“thinks” in order to be able to properly interpret the
results. The way the model arrives at those results
has critical implications for the selection of data.

Scatter graph

Do you recall what a regression residual is? It is the


“error” or difference between what the regression
model (or graph) would predict as the dependent
variable and the actual value of the dependent
variable. On your graph it is the space between each
“dot” on your scatter graph and the regression line
you drew that represents the average relationship
between size and price. With that in mind, the rule
the regression model uses to pick its line is called the
rule of least squared errors. The best way to explain
the rule is by example. Look back at your graph. If
you calculated the difference between the regression
line and the actual selling price for each of the
transactions in the sample, then squared each of
those differences and added them up, you would
have the sum of the squared errors for that line. Now
suppose you redrew your line slightly differently and
once again got the sum of the least squared errors for
that line. Now suppose you did this an infinite number
of times, trying every possible line, and calculating
the squared error for each. The line the regression
model selects is the line that produces the lowest
sum of the squared errors—thus the term least
squared errors.

1 of 2 4/5/2012 7:13 AM
AI http://onlineed.appraisalinstitute.org/ai/OnlineEd/Onlineed.asp?Session_...

Page 18 of 22

The Appraisal Institute advocates equal opportunity and nondiscrimination in the appraisal profession
and conducts its activities in accordance with federal, state and local laws.
©2012 Appraisal Institute. All rights reserved.
Terms of Web site use.

2 of 2 4/5/2012 7:13 AM

You might also like