ST3131: Regression Analysis
Key Points for Revision
1. The variants of multiple regression models. Dummy variables;
interpretation of regression parameters: main-effect model, interaction
model, and covariance analysis models.
2. Multiple comparison. Contrasts in terms of means and in terms of
regression parameters; The three criteria: Sheffe, Tukey and Bonfer-
roni; The respective situations where the criteria are applicable; The
relations among the three criteria; The application of the criteria in
two-sided and one-sided tests; Simultaneous confidence intervals; p-
values of tests for multiple comparison.
3. Variable selection. Variable selection procedures; Details of the se-
lection procedures at each step; Selection criteria; Computation of AIC
and BIC.
4. Model diagnostics. Raw materials for diagnostics and their respec-
tive usages; How to check for appropriateness of regression function;
How to check for non-constancy of variance; How to check for normal-
ity; How to identify outliers? The type of outliers; The null patterns
in various diagnostic plots, etc.
5. Remedial measures. Weighted least square estimation and the deter-
mination of weights; Methods for detecting multicollinearity; Variance
inflation factor; Variance stabilization transformation.
Note: The final exam will place emphases on the above points. But some
basic aspects of regression models might also be covered such as one-sided
and two-sided tests, tests for linear hypothesis, R2 and multiple correlation
coefficients.
1



























学霸联盟