665 REGRESSION

DEPARTMENT OF STATIS TICS
COLLEGE OF ARTS & SCIENCES
330.972.6886 (TELEPHONE)
Course Description
3470:665
Prerequisite:
Regression
3 credits
3470:461/561 Applied Statistics or equivalent.
Course Description: Correlation, simple and multiple linear regression, least
squares, matrix notation, model building and checking estimation, hypothesis testing,
outliers, influence, multicollinearity, transformations, categorical regressors, logistic
regression.
Course Topics:
Introduction to regression models and their applications
Simple Linear Regression
SLR models
Lease squares
Estimation, prediction, ANOVA, and hypothesis testing in SLR
Checking model assumptions, including residual analysis
Issues: regression to the mean, extrapolation, causality
Correlation Analysis
Pearson’s linear coefficient
Spearman’s rank correlation coefficient
Multiple Linear Regression
MLR model – matrix formulation
Least squares
Estimation, prediction, ANOVA, and basic hypothesis testing in MLR
Sequential and partial sums of squares
Testing of sets of regressors (reduction in sums of squares principle)
Modeling Checking
Residual plots
Identifying outliers (studentized residuals)
Leverage (hat matrix)
Influence diagnosis
Diagnosing multicollinearity
Lack of fit tests
Check for autocorrelation
Cross-validation (and PRESS)
Model Building
Transformations
Variable selection criteria (e.g., Cp statistic)
All possible regressions and stagewise procedures
Inclusion of Categorical Regressors in the Model
Weighted Least Squares
Alternative Regression Approaches (Optical topics)
Ridge regression
Principal components regression
Nonlinear regression
Nonparametric regression
Path analysis
Fall 2014