750 / 762
750
Chapter 16: One-Way Analysis of Variance
We can now write up the hypothesis test for independence. Note the number
of errors was 37. This is the value of
.
: The errors are independent of each other.
: The errors are not independent of each other.
: |
| = | -0.1 | = 0.1.
: Reject the null hypothesis if |
| = 0.1 > 1.96)
(
√ ⁄
). That
is, reject the null hypothesis if 0.1 > 0.3222.
Conclusion
: Since 0.1 < 0.3222, do not reject the null hypothesis.
That is, there is insufficient sample evidence to conclude that the errors are
not independent of each other. Thus, one may assume that the independent
assumption for the errors has not been violated.
Validating the Constant Variance Assumption for the Model
Errors
When we can assume normality for the distribution of the errors, there is a
test which we can use to test for equal variance for several populations. This
test is called the
Bartlett’s Test for Equality of Variances
. The Levene’s
test assumes that the error distribution is continuous.
If we cannot assume normality for the errors, we can use the
Levene’s test
for Equality of Variances.
Both of these tests are in the
Test for Equal Variance
workbook.
We will use the Bartlett’s test to test for the constant variance assumption
using a level of significance of
= 0.01 since we can assume normality.
A partial output of the results is displayed in
Figure 16-27
.
Click here for the Auto Correlation Function Workbook