SFOK18003U Statistical analysis of repeated measurements with mixed models

Volume 2018/2019
Education

MSc in Public Health Science - elective course

This course is also open for students in the following programmes, but please be aware of the recommended academic qualifications:

MSc in Global Health

MSc in Health Informatics

MSc in Human Biology

MSc in Health Science

 

Content

This course is concerned with analysis of correlated quantitative data arising
e.g. when taking obsertions repeatedly over time on the same subjects or from clusters of subjects such as siblings, families, school classes or patients belonging to the same cilic. Pitfalls of traditional statistical analyses will be
discussed and appropriate models for the analysis of e.g. baseline follow-up studies, cross-over trials, and cluster randomized trials will be exemplified.

Day 1: Basic concepts for repeated measurements. Descriptive
statistics. Baseline follow-up studies. Linear mixed models. Handling repeated measurements in SAS or R.

Day 2: Longitudinal data analysis. Modeling mean and covariance. Balanced and unbalanced data.

Day 3: Analysis of clustered data. Variance components. Multi-level models. The linear growth model.

Day 4: Select applications of linear mixed models. Cross-over trials.
Repeatability and reproducibility of measurement methods

Day 5: Models for binary and other non-normally distributed data. Generalized linear mixed models. Marginal models and generalized estimating equations.

Day 6: Missing data. Consequences and statistical methods for handling. Course summary and evaluation.

 

Learning Outcome

After completing the course, the students are expected to be able to:

Knowledge

Have a working knowledge of statistical methods for analysing data from repeated measurements and clustered data designs including baseline-follow-up studies, cross over trials, studies of reproducibility, cluster randomized tirals, and multilevel models.

Have a working knowledge about linear mixed models for quantitative outcomes as well as generalized linear models and generalized estimating equation for non-normally distributed outcomes.

Reflect on the statistical consequences of using repeated measurements or clustered designs compared to independent samples designs.

Reflect on potential biases due to missing data in studies with repeated measurements or clustered data.

Skills

Perform analyses of repeated measurement and clustered data including baseline-follow-up studies, cross over trials, studies of reproducibility, cluster randomized tirals, and multilevel models using SAS or other statistical software.

Interpret the results of the above-menationed analyses of repeated measurements or clustered data.

Assess the validity and potential limitations of the results from the above-mentioned analyses of repeated measurements or clustered data.

Present results from analyses of repeated measurements or clustered data in figures, numbers, and text.

Read and discuss statistical results from analysis of repeated measurements or clustered data presented in the health science literature.

Competences

Independently plan and conduct analyses of repeated measurements or clustered data as part of your own master, ph.d., or other health science project.

Collaborate with a statistician in planning and conducting analyses of repeated measurements or clustered dat as part of your own master, ph.d., or other health science project.

Take part in dicussing results of health science research that is based on analysis of repeated measurements or clustered data.

Text book: G.M. Fitzmaurice, N.M. Laird and J.H. Ware, Applied Longitudinal Analysis (2nd edition), John Wiley & sons, 2011.

Please note that an e-book version can be downloaded from the Royal Library.

Lecture notes, exercise problems, datasets and other supplementary material will be made available at the course webpage

http:/​/​publicifsv.sund.ku.dk/​~jufo/​RepeatedMeasures2018.html

as the course proceeds.

Students must have completed a course on basic statistics so that they have a solid working knowledge of elementary statitical models for independent normally distributed data such as the two-sample and paired t-tests, one- and two-way analysis of variance, linear regression and correlation, and the general linear model. Moreover familiarity with logistic regression/generalized linear models is a requirement.
Finally students must be familiar with either SAS or R programming to be able to solve the exercise problems.
Forum lectures and exercise classes with either SAS or R programming. IMPORTANT: All students are required to bring their own labtops with either SAS or R installed (or a license to SAS Studio).This is essential for working with the exercise problems.
  • Category
  • Hours
  • Class Instruction
  • 18
  • Exam
  • 56
  • Lectures
  • 18
  • Preparation
  • 46
  • Total
  • 138
Oral
Individual
Collective
Continuous feedback during the course of the semester

All students may receive personal supervision from the teachers during the exercise classes. You are encouraged to engage in discussion with the teachers and your fellow students in particular during exercise classes, but questions during lectures are also welcome.

Students planning to analyse repeated measurements data in their master or ph.d. project are asked to reflect on their research project in class.

Credit
5 ECTS
Type of assessment
Written assignment, 1 week
Written take home exam. Students may discuss the exam problems but are requested to hand in individual reports.
Aid
All aids allowed
Marking scale
passed/not passed
Censorship form
No external censorship
Ingen censur.
Exam period

Please see the exam schedule at KUnet 

Re-exam

Please see the exam schedule at KUnet 

The exam form in the re-examination may differ from the ordinary exam. Should this happen, students registered for the re-examination will be informed as soon as possible.

Criteria for exam assesment

To pass the exam, the students are expected to be able to:

Knowledge

Have a working knowledge of statistical methods for analysing data from repeated measurements and clustered data designs including baseline-follow-up studies, cross over trials, studies of reproducibility, cluster randomized tirals, and multilevel models.

Have a working knowledge about linear mixed models for quantitative outcomes as well as generalized linear models and generalized estimating equation for non-normally distributed outcomes.

Reflect on the statistical consequences of using repeated measurements or clustered designs compared to independent samples designs.

Reflect on potential biases due to missing data in studies with repeated measurements or clustered data.

Skills

Perform analyses of repeated measurement and clustered data including baseline-follow-up studies, cross over trials, studies of reproducibility, cluster randomized tirals, and multilevel models using SAS or other statistical software.

Interpret the results of the above-menationed analyses of repeated measurements or clustered data.

Assess the validity and potential limitations of the results from the above-mentioned analyses of repeated measurements or clustered data.

Present results from analyses of repeated measurements or clustered data in figures, numbers, and text.

Competences