Document Sample
Announce Powered By Docstoc
                                Starting Spring 2009

                STAT 4510/7510: Applied Statistical Models I
                  STAT 8220: Applied Statistical Models II

               A fresh update our classic Regression and ANOVA courses

Stat 4510/7510, formerly called Regression and Correlation Analysis, has been updated
and re-titled Applied Statistical Models I. The content of this course has been revised to
provide a unified presentation of linear models including classical regression and analysis
of variance topics. By presenting classical regression and ANOVA topics in a cohesive,
condensed format, students will also be exposed to random and mixed effects models in
the first semester. Topics in the “new” 4510/7510 include simple linear regression,
multiple regression, analysis of designed experiments, multi-factor ANOVA, random
effects models, and mixed effects models.

Starting in FS2009 the second course in the new sequence, Stat 8220: Applied Statistical
Models II, will be offered. This course will focus on important modern modeling
techniques with an emphasis on methods and procedures that are of interest to applied
researchers. Topics include linear mixed models, generalized linear models, and
nonlinear models. Optional topics, time permitting, are generalized linear mixed models,
Bayesian estimation, longitudinal data models, nonparametric regression, generalized
additive models and log-linear models.

During both semesters, students will be expected to apply the techniques learned in class
to datasets using statistical software. We anticipate the use of SAS during the first
semester SAS and/or R during the second semester.

At the present time, we plan to continue offering Stat 4530/7530: Analysis of Variance.
Depending on student need, we will re-evaluate the necessity of continuing this course in
the future.

Other statistics courses that students may find important in meeting their career goals
may be found in our Schedule of Planned Statistics Offerings, which is a useful tool for
curriculum planning.

                      Stat 4510/7510: Applied Statistical Models I
           To be offered each fall, spring, and summer starting in Spring 2009

Catalog description: Introduction to applied linear models including regression (simple
and multiple, subset selection, estimation and testing) and analysis of variance (fixed and
random effects, multifactor models, contrasts, multiple testing). No credit for a graduate
degree in statistics. Prerequisite: STAT 3500 or 7070 or 4710/7710 or 4760/7760 or
instructor’s consent.

Additional requirements for graduate credit: Graduate students will be required to
analyze a data set of their own choosing and, time permitting, to make a presentation on
their analysis to class.

Possible textbook:
Applied Regression Analysis and Other Multivariable Methods, 4th Edition, by
Kleinbaum, Kupper, Nizam and Muller.

                                     Course Topics:

Review (1 week)
      Descriptive statistics; one sample inference for mean and variance, Type I and II
      errors, sample size and power; two-sample comparison of means and variances.
Simple linear regression (3 weeks)
      Basic regression model and assumptions, least squares estimation; normal error
      and inference on regression coefficient, intercept, mean response, prediction;
      ANOVA table, residuals and regression diagnostics, measurement errors.
Multiple Regression (4 weeks)
      ANOVA table, inference on regression coefficient, mean and prediction, partial
      determination, Type I and II sums square, polynomial regression, categorical
      variables, interaction term, diagnostic (outliers of X, Y, influential observation,
      multicolinearity), remedial measures (weighted least squares, variance stabilizing
      transformations), model selection and validation.
Analysis of Designed Experiments / One Way ANOVA (1 week)
      Completely randomized designs, one-factor fixed effects model, alternative
      formulations and restrictions on parameters, contrasts, multiple comparisons;
      estimation and inference, F- and t-tests.
Multi-Factor ANOVA (3 weeks)
      Two-factor fixed effects models (balanced and unbalanced designs, alternative
      models formulations and restrictions, interaction, orthogonal contrasts, non-
      orthogonal decompositions, Type III sums of squares, F-tests.
Introduction to Random Effects (1 week)
      One-factor random effects models, estimation, F-tests; two-factor random/mixed
      effects models (balanced designs, variance components, estimation and
Analysis of Covariance (1 week)
Software: SAS

                       STAT 8220: Applied Statistical Models II
            To be offered each fall, spring, and summer starting in Fall 2009

Course description: Advanced applied linear models including mixed linear mixed
models (fixed and random effects, variance components, correlated errors, split-plot
designs, repeated measures, heterogeneous variance), generalized linear models (logistic
and Poisson regression), nonlinear regression. No credit for a graduate degree in
statistics. Prerequisites: STAT 4510/7510 or instructor’s consent.

Possible Textbooks:
       Applied Regression Analysis and Other Multivariable Methods, 4th Edition, by
       Kleinbaum, Kupper, Nizam and Muller.

       Linear Mixed Models by West, Welch and Galecki

                                     Course Topics:

Review of Fixed and Random Effects models

Linear Mixed models
      Hierarchical framework, correlated error structure, split-plot designs, repeated
      measure designs, heterogeneous variance.

Generalized Linear Models
      Logistic regression (binomial response), Poisson regression, model structure (link
      functions), missing values, inference, overdispersion

Nonlinear Regression Methods
      Nonlinear least squares, numerical methods, practical issues, statistical properties,

Additional topics (time permitting):
       Generalized linear mixed models, Bayesian estimation, longitudinal data,
       nonparametric regression, generalized additive models, log-linear models

Software: R, SAS, etc.