Section 11.5 Inferences Concerning the Regression Coefficients
Confidence Interval for b
, the slope parameter
A (1-a)100% confidence interval for the parameter b in the regression line mY|x = a + bx is
where is a value of the t-distribution with n-2 degrees of freedom
Hypothesis testing on the slope parameter, b
To test the null hypothesis Ho:b = bo against any suitable alternatives,
use t-distribution with n-2 degrees of freedom to define the critical region, and the following test statistic
Statistical Inference on the intercept
Confidence Interval for a
, the intercept parameter
A (1-a)100% confidence interval for the parameter a in the regression line mY|x = a + bx is
or
where is a value of the t-distribution with n-2 degrees of freedom
Hypothesis testing on the intercept parameter, a
To test the null hypothesis Ho: a = ao against any suitable alternatives,
use t-distribution with n-2 degrees of freedom to define the critical region, and the following test statistic
df=33-2=31, P-value =
From Table A.4 directly, P-value < 0.05. So, the null hypothesis of zero intercept is rejected at 0.05 level of significance.
A measure of quality of fit: Coefficient of
Determination, R2
Coefficient of Determination, R2 = proportion of total variability in the dependent variable Y explained by the fitted model.
TSS = SSR + SSE
Coefficient of Determination,
R2 = 1.0 if fit is perfect
R2 = 0.0 if fit is poor
Section 11.6 Prediction
Confidence Interval for , the mean response
A (1-a)100% confidence interval for the mean response in the regression line mY|x = a + bx is
where is a value of the t-distribution with n-2 degrees of freedom
Prediction Interval for , the future value of a response
A (1-a)100% prediction interval for a single response in the regression line mY|x = a + bx is
where is a value of the t-distribution with n-2 degrees of freedom
Section 11.11 Simple Linear Regression
Case Study
A more complicated model may be more appropriate
The higher R2 value would suggest that the transformed model is more appropriate.
Section 11.12
Correlation
Correlation coefficient
The measure r of linear association between two variables X and Y is estimated by the sample correlation coefficient r, where
-1 < r < 1
r = 1.0 or –1.0 perfect linear relationship
r = 0.0 no linear relationship
Sample coefficient of determination
represents the proportion of the variation in TSS explained by the regression of Y on x, namely, SSR.
Hypothesis testing on the correlation coefficient, r
To test the null hypothesis Ho: r = ro against any suitable alternatives,
We can use t-distribution with n-2 degrees of freedom to define the critical region, and the following test statistic
but this test has problem when r values close to -1 or 1
However, a more general test is given by
the approximate normal distribution with mean and variance .
We can use the standard normal to define the critical region, and the following test statistic
Important: Correlation is a measure of linear relationship, so r = 0 does not necessarily mean there is no relationship between two variables.