This course explores methods for the analysis of longitudinal data and latent variable methods for linear models, generalized linear models, and nonlinear models.  Focusing on applications, this course explores: the analysis of repeated measures ANOVA, multivariate approaches, random-effects regression, covariance-pattern models, generalized-estimating equations and generalizations, latent variable methods including finite mixture modelling, and likelihood methods.  Students will develop expertise using the SAS and R computer packages, although no previous programming experience will be assumed.  Grading is based on homework and computer assignments and a project, as well as several exams.

Course begins with a lecture on Monday 17th of September at 2 pm.

There are three exams during the course.

Recommended preceding studies:

Basic courses of statistics and Regression analysis

Content

It is quite common that we do not get all the information we want for our statistical analysis. For example, in medical research a person can refuse to provide certain information they feel sensitive, such as weight, substance abuse, sexual orientation etc. Particularly, missing data in longitudinal studies is more the rule than an exception. Missing data in statistical analysis causes all sorts of problems. For example, the desired statistical method cannot be directly applied; loss of information or the results obtained can be biased if the analysis is not done properly accomplished. The course introduces various missing data mechanisms and their effects on statistical analysis. In addition, it presents and evaluates some of the commonly used methods for statistical analysis with missing data. Also special methods for the analysis of longitudinal data are presented including likelihood-based methods and multiple imputation.

Modes of Study

Course work, exam.

This course is an intermediate study version of MTTS1 Statistical Analysis with Missing Data. The student can only complete one of the two versions.

This course will give a detailed overview of statistical models for modern regression and classification with emphasis on applications. A number of examples and case studies will be examined. This course will cover a range of models from linear regression through various classes of more flexible models including fully nonparametric regression models. We will consider both regression and classification problems. Methods such as splines, additive models, multivariate adaptive regression splines (MARS), neural networks, classification and regression trees (CART), linear and flexible discriminant analysis, generalized additive models, nearest- neighbor rules and learning vector quantization will be discussed.

Recommended preceding studies:

Basic courses of statistics and Regression analysis.

Please note

Students who have completed course MTTA2 Ei-parametrinen regressio can not get full credits of this course because some of the contents overlap.