KLMED8008: Analysis of repeated measurements
This
course is aimed at PhD students of medicine and health sciences.
Language
The lectures, textbook
and the exercises are in English.
Contents and credits
Most commonly employed
statistical procedures require that observations are independent. In this
course we will discuss problems, advantages, and strategies for
analysis of repeated measurements on the same
individual or group.
The course gives 5 study credits (studiepoeng).
Prerequisites
(Forkunnskaper)
Basically, this
course builds on KLMED8005. Knowledge of and experience in multiple linear regression will prove very helpful.
Course supervisor
Professor Eirik Skogvoll, MD PhD, Dept. of Anaesthesia and Emergency Medicine, St.Olav University Hospital
Phone +47 725 74549
e-mail: eirik.skogvoll@ntnu.no
Textbook
Rabe-Heskett & Skrondal: Multilevel and Longitudinal Modeling Using Stata, Stata Press, 3rd ed. (2012), Volume I: Continuous Responses.
We will cover chapters 2 to 5 with some exceptions. Supplementary material from the lectures.
(One may also use the second edition of this book (2008), but since then some Stata code and commands have changed).
Software
Basically, we will follow the textbook and use Stata.
It will therefore be very useful to run
Stata from your own laptop during the course. However, you may use your
favourite software to solve the exercises and the exam as long as the
results are interpreted and presented clearly.
Stata should be available for NTNU users. The current release is
Stata 12.
You may run Stata over the network from your own computer, or from Terminal Server (TS). N
avigate to
P:\begrenset\STATA-IC and double-click on the Stata12 icon.
Disregard any start-up messages about "update", this is done automatically.
It will be useful to create a shortcut (see below).
Data files and do-files
Data files in Stata format (*.dta) and r
elevant do-files (Stata command files) that come with the book may be downloaded from the Textbook's website:
http://www.stata-press.com/books/multilevel-longitudinal-modeling-stata
Installing add-ons for Stata
If you would like to install add-ons (which may be very useful; e.g.
glamm), it is necessary to tell Stata where the program should save the
files.
You need to create two directories under your home catalog (the full
path is shown here, and please substitute your username); e.g.
\\home.ansatt.ntnu.no\your_username\stata-data\personal
\\home.ansatt.ntnu.no\your_username\stata-data\ado
and then run the following commands either via a do-file or from the Stata prompt:
sysdir set PERSONAL
\\home.ansatt.ntnu.no\your_username\stata-data\personal
sysdir set PLUS \\home.ansatt.ntnu.no\your_username\stata-data\ado
This may be done automatically at startup if the abowe two commands are
found in a file named "profile.do", located in the Stata working
directory.
A
simple way to ensure that Stata starts in the appropriate directory is
to create a shortcut on your computer or TS desktop. Choose shortcut
properties, and specify that Stata should start in the directory
where "profile.do" is found, e.g.
\\home.ansatt.ntnu.no\your_username\stata-data
Enrolment
(Påmelding)
Via StudentWeb. Complete this before study start. Closing date is 15.02.2013.
For students not registrered at NTNU (For studenter uten studierett ved NTNU): please fill in the relevant application form: http://www.ntnu.no/dmf/forskning/phd/emner
Course administrator
Hege Müller Hval
Phone +47 (728) 25263
Visiting address: Prinsesse Kristinasgate 1 (Gastro centre)
e-mail:
hege.m.hval@ntnu.no
Course schedule as of May 29.th 2013.
Time: Wednesdays at 1215 - 1400 in LS41 mostly, or LS42. See below.
LS41 and LS42 - Seminar rooms, 4th. floor, Laboratory centre
|
Week no.
|
Subject
|
Date and place
|
Lecturer
|
Literature*
|
9
|
Day 1. Introduction.
Problems and advantages with dependent observations.
Review of variance, standard deviation, covariance and correlation.
Brief introduction to Stata.
Summary measures (mean, median, area under curve (AUC), slope coefficient, min/ max value)
Review of multiple linear regression.
|
27.02.2013
LS41
1215-1400
|
Eirik Skogvoll
|
Slides (corrected)
Do-file
Rabe-Heskett & Skrondal: Chapter 1
Matthews et al. (1990)
Bland & Altman (1995)
Rosner_6 Section 5.6: covariance and correlation
Data sets (Stata format):
X67 X67s gtube gtube_l
|
10
|
Day 2.
Comparing groups with adjustment for baseline using ANCOVA.
Power and sample size determination.
Brief discussion of ANOVA, repeated measurements ANOVA and Friedman's non-parametric test.
|
06.03.2013
LS41
1215-1400
|
Eirik Skogvoll
|
Slides
Do-file
Vickers & Altman (2001)
Kinnear & Gray: Chapter 9
Rosner_6: section 12.6
Data sets (Stata format):
Vickers lwiner winer wwiner
|
11
|
Day 3.
Robust variance estimation in linear regression.
Review of statistical models in general.
Variance components.
|
13.03.2013
LS41
1215-1400
|
Eirik Skogvoll
|
Slides
Do-file
Rabe-Heskett & Skrondal: Chapter 2
Rosner_6: sections 12.8, 12.9
|
12
|
Day 4.
Excercise no. 1. Voluntary, no hand-in. Lecturer present.
|
20.03.2013
LS41
1215-1400
|
Eirik Skogvoll
|
Exercise 1
|
13
|
Day 5.
Review of exercise no 1.
Power and sample size determination with clustered design.
Linear mixed effect models.
Random intercept models with covariates.
|
03.04.2013
LS42
1215-1400
|
Eirik Skogvoll |
Slides (revised)
Rabe-Heskett & Skrondal: Chapter 3
Gueorguieva,
R. and J. H. Krystal (2004).
Stata manual [XT] (pdf, comes with the program)
|
15
|
Day 6.
Random coefficient models.
Brief discussion of Generalized Estimating Equations (GEE).
|
10.04.2013
LS41
1215-1400
|
Stian Lydersen
|
Notes
Rabe-Heskett & Skrondal: Chapter 4 |
16
|
Day 7.
Excercise no. 2. Voluntary, no hand-in. Lecturer present.
|
17.04.2013
LS41
1215-1400
|
Øyvind
Salvesen
|
Do-file1
Do-file2
Exercise 2
|
17
|
Day 8.
Review of exercise 2.
Longitudinal models.
|
24.04.2013
LS41
1215-1400
|
Øyvind Salvesen
|
Rabe-Heskett & Skrondal: Chapter 5
Slides
Do-file1
|
18
|
Day 9.
Longitudinal models (cont’d).
Power and sample size determination.
Presentation and discussion of students' own projects.
Summing up, questions.
|
08.05.2013
LS41
1215-1500
|
Eirik Skogvoll
Øyvind
Salvesen
|
Rabe-Heskett & Skrondal: Chapter 6
Do-file1
Do-file2
Do-file3
|
Suggested literature in addition to the textbook:
- Matthews, J. N., D. G. Altman, et al. (1990). Analysis of serial measurements in medical research. BMJ 300: 230-235.
- Bland JM, Altman DG. Calculating
correlation coefficients with repeated observations: Part 1- correlation within
subjects. BMJ 1995;310:446
- Rosner, B. Fundamentals of Biostatistics (5ht, 6th or 7th edition). Relevant sections on variance, standard deviation, covariance, correlation, and variance components.
- Gueorguieva,
R. and J. H. Krystal (2004). Move over ANOVA: progress in analyzing
repeated-measures data and its reflection in papers published in the
Archives of General Psychiatry. Arch Gen Psychiatry 61(3): 310-7
- Kinnear P, Gray C: SPSS 14 Made Simple (2006) Psychology Press
- Vickers,
A. J. and D. G. Altman (2001). Statistics notes: Analysing controlled
trials with baseline and follow up measurements. BMJ 323: 1123-4.
- Veierød, Lydersen
& Laake (ed.) Medical Statistics in clinical and epidemiological
research. Chapters 7 and 8. Oslo: Gyldendal 2012.
Exam
A "take home" exam (including a dataset) will
be handed out on May 29.th between 0900 and 1200 hours. Please turn up in person at the Department
of Cancer Research and Molecular Medicine, Gastro centre, 3rd floor,
St. Olav Hospital. Press the doorbell if you can't get in! The dataset to be analysed is found here: asthma2.dta (Stata) asthma2.dat (tab-delimited) asthma2.sav (SPSS). Contact person: course administrator Hege Müller Hval, phone +47 (728) 25263.
The exam is to be analyzed and returned (same place) within one week, on June 5.th, between 0900 and 1200 hours. Good luck!