Coefficient of Determination Section 4.3 Alan Craig 770-274-5242 [email protected].
-
Upload
dora-maxwell -
Category
Documents
-
view
221 -
download
3
Transcript of Coefficient of Determination Section 4.3 Alan Craig 770-274-5242 [email protected].
2
Objectives 4.3
1. Compute and interpret the coefficient of determination.
3
Coefficient of Determination
• The coefficient of determination, R2, measures the percentage of the total variation in the response variable that is explained by the least-squares regression.
• R2 is calculated by squaring the linear correlation coefficient, r.
4
On the Calculator
r and R2 are part of the calculator output for a linear regression with DiagnosticsOn.
5
Least-Squares Regression
• Recall that the least-squares regression line minimizes the sum of the squared
errors (residuals) = residuals2
Error or residual = actual - predicted
6
Estimating y
If I have no information about values of the predictor variable x, then my best guess for y is the mean of y:
and the deviation is the actual value minus the mean.
y
7
Actual
Deviation
yMean
yy
Total Deviation
y
8
Estimating y
However, if I have additional data on the values of x and corresponding values of y, I can often do better by calculating the regression of y on x.
Part of the total deviation is now explained by the regression equation although some of the deviation is still unexplained (unless there is a perfect linear correlation).
9
Actual
Deviation
yMean
Predictedyy ˆ
yy ˆ
Unexplained Deviation
Deviation explained
by the regression
y
y
10
Deviation
Note that
That is,
y = mean of y
+ explained deviation
+ unexplained deviation
)ˆ()ˆ( yyyyyy
11
Deviation
Or
That is,
Total deviation = explained deviation
+ unexplained deviation
)ˆ()ˆ( yyyyyy
12
Variation
The total variation of y is
The explained variation is
The unexplained variation is
1
2
n
yy
1
ˆ 2
n
yy
1
ˆ 2
n
yy
13
Variation
variationtotal
variationdunexplaine1
variationtotal
variationexplained
variationtotal
variationdunexplaine
variationtotal
variationexplained1
variationdunexplaine variationexplained variationtotal
R2
14
Interpreting R2
Thus, R2 is the percentage of variation in the response variable, y, that is explained by the predictor variable x.
variationtotal
variationdunexplaine1
variationtotal
variationexplained2 R
15
Interpreting R2
Using our example from Sections 4.1 and 4.2 (problem 10, p. 172), R2 =0.9835=98.35%, so the predictor variable, Carats, explains 98.35% of the variation in the response variable, Price.
16
Questions
• ???????????????