Courses that I have taught at Michigan
Service Courses
Biostatistics 523
The goal of this course is to provide students with skills necessary to
carry out regression analyses, commonly used method in many health
investigations. Both experimental designs and
observational studies are considered. The regression models considered are linear,
nonlinear, logistic, Poisson, multinomial and ordinal regression models
with an emphasis on multiple linear regression models. The
Analysis of Variance (ANOVA) and covariance (ANCOVA) are also be
covered. The ideas of confounding, effect modifications/interactions,
and causal versus association inferences are also covered. There are
extensive discussions of inferences and interpretations, approaches for
fitting and assessing goodness of fit of these regression models. The
primary software package used in this class is SAS. The Lab
component of the course consists of practical implementation of
the methods covered in the lectures. This is a second course in
biostatistics and the first course in regression analysis.
Biostatistics 524
The goal of this course is to provide students basic understanding of
statistics. The topics include graphical and numerical presentation of
data, Design of Experimental and Observational Studies, Probability,
Basics of inference: Estimation, sampling distribution, confidence
intervals and signifcance tests. Numerous applications one sample
and two sample problems, continuous and discrete outcomes.
Nonparametric methods and simple linear regression. This course is a
part of the University of Michigan's Master's degree in Clinical
Research Design and Statistical Analysis (CRDSA). This is an 18-month
program where students meet one extended weekend (Thursday-Sunday) a
month to take courses and do homework problems between sessions. This
program allows students to maintain full time employment and earn
Masters degree. For more details see OJOC.
Biostatistics 560
The objective of this course is to learn fairly advanced statistical
techniques used in Epidemiology. This course begins with logistic and
Poisson regression models, survival analysis, log-linear models for
categorical data, models for longitudinal data. Students use SAS to
analyze several data sets. This course requires Biostatistics 523. Most
of the students are doctoral students in Epidemiology.
Biostatistics 581
The objective of this course is to learn about statistical models and
methods for the analysis longitudinal data. The emphasis of this course
is on applications in health sciences. The basic theoretical
reasoning underpinning each method/model are be provided. In this
course the students learn different parametric approaches for
analyzing repeated measurements and explore the assumptions underlying
each approach. Issues concerning missing data in longitudinal studies
are addressed. The student learn several aspects of
analyses of longitudinal data with continuous, binary, ploytomous and
count outcome variables in greater depth through several data analysis
examples and term projects. This course is also a part of the OJOC
program.
Courses taken by Masters and Doctoral Students in Biostatistics
Biostatistics 650
This is an introductory regression analysis with emphasis on Multiple
linear regression. This course is taken by the first year students in
the Biostatistics Masters program. All aspects of ordinary least
squares, weighted, ridge and robust regressions are discussed. Model
diagnostics and remedies are discussed. All aspects of inferences based
on these models are also discussed. Nonlinear and generalized linear
regreession models are briefly introduced. Students use R and SAS to
peform analysis of data.
Biostatistics 682
This is an introductory Bayesian analysis courses for masters and
doctoral students. This course is taken by students in Biostatistics,
Statistics and Engineering. The students are introduced basics of a
Bayesian view of statistical inference, simple examples and
introduce Bayesian computations. The details about various iterative
and non-iterative simulation techniques, model checking and diagnostics
and several practical applications are discussed. The students
use R, SAS and Winbugs to perform analysis.
Biostatistics 870
This course is designed for doctoral students in Biostatistics and
Statistics and covers both theoretical and practical aspects of
modeling longitudinal data. Population average and subject specific
models are discussed; Both parametric and nonparametric models are
discussed; Estimating equations, maximum likelihood and Bayesian
approaches are discussed. Missing data issues are also dicussed.
Students write their own programs in R, SAS, C or Fortran to develop
and evaluate methodology.
Biostatistics 880
This course is taken mostly by doctoral and by some masters level
students and covers the topic of analysis of incomplete data. The
topics include Missing data mechanisms and patterns, possibe approaches
for analysing data with missing values: Weighting, multiple imputation,
maximum likelihood and fully Bayesian approaches. Students learn both
theoretical and practical aspects of inference from incomplete data
under a variety of contexts: Experimental design, observational
studies, cross-sectional and longitudinal data and ignoorable and
non-ignorable missing data mechanism.
Survey Methodology Courses
SURVMETH 612
This is an applied sampling course taken by the Masters and doctoral
level students learn about design and estimation problems from complex
surveys. More emphasis is on designing of surveys including telephone,
mail, web and face-to-face surveys. Frame construction, inclusion
probability calculations, clustering, stratification and weighting are
dicussed. Approximations needed to calculate the sampling variances of
population means, proportions and totals are discussed. Area
probability sampling and dual frame designs are also discussed.
SURVMETH 613
The objective of this course is to introduce advanced analytical
technique using the complex survey data. Inference for parameters in
various regression models are discussed using linearization and
replication techniques. Students learn, through practical
examples, the impact of ignoring the complex design features.
Approximations of designs to calculate the sampling variances are
discussed. Weighting and imputation methods for handling incomplete
data are dicussed.
SURVMETH 895-1
The objective of this course is to provide theoretical and practical
aspects of weighting and imputation in complex surveys. Cell
adjustment, response propensity and post-stratification methods for
constructing weights are discussed. Both implicit (such as Hotdeck) and
explicit model based imputation methods are discussed. Practical
examples are used to illustrate the methodology and get hands-on
experience.
SURVMETH 895-2
The objective of this course is to introduce Bayesian Inference for
Surveys. The emphasis is on inferring abou the finite population
quantities through predictive inference of the non-sampled portion of
the population. All complex design features are discussed in the
Bayesian context including ignorable and non-ignorable sampling
mechanisms. The methodolgy is implemented through writing
programs in R, SAS and Winbugs.