SCS 2017: Longitudinal and Nested Data/Course description

From Wiki1

Jump to: navigation, search

Instructor

Have a look at Piazza for some information

Description

The course focuses on models and methods suitable for longitudinal data in which each subject is observed on a number of occasions over time. Each subject may be observed a different number of times and times may be spaced differently for different subjects. In contrast with classical repeated measures designs, the methods we consider handle unbalanced data, including time-varying covariates.

These methods are also appropriate for nested data with no time component but observations are clustered in groups.

The course will begin by developing concepts and techniques that are specifically relevant for nested and longitudinal data: random versus fixed effects, variance-covariance components, temporal autoregression, contextual versus compositional effects, splines, missing data and diagnostics, among others.

We will apply these concepts with mixed models using software in the statistical programming environment R (https://cran.r-project.org/), including the ‘nlme’ package for linear and non-linear modelling with continuous responses.

We will then learn Bayesian methods using Markov Chain Monte Carlo techniques that can be used for more complex models and for discrete responses. For this purpose we will focus on the Stan modelling language and program (http://mc-stan.org/).

This short course assumes familiarity with linear regression as presented, for example, in John Fox, Applied Regression Analysis and Generalized Linear Models, Third Edition (Sage, 2015). Familiarity with the basics of R will also be an asset and participants are encouraged to install R and work through an introductory tutorial, such as the one at ( https://cran.r-project.org/doc/manuals/r-release/R-intro.html ) to prepare for the course. Another option is Swirl: a package in R with interactive tutorials for different skill levels. You need to know how to run R and install packages, but there are instructions on the site — http://swirlstats.com/ . If participants have completed the first course in this series (see above) they will be well prepared. In this course, extensive examples will be given in the R programming environment.

Because this material is presented sequentially and builds upon the basics presented at the beginning of each class, course participants need to arrive on time and attend the entire session.

You are encouraged to bring your laptop. There will be many opportunities to practise using examples in R. Wireless access will be provided for non-York community participants.

Personal tools