To page body

Logistic Regression Analysis Course

Time: Week 12    21.-24.3.2016

Volume: 2 ECTS

Place: School of Health Sciences, Building T, University of Tampere and Computer class ML71 Arvo building

Lecturer: Professor Stephen Walter, Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, Canada

Coordinator: Lecturer Anna-Maija Koivisto, e-mail: anna.m.koivisto(at)

Organisers: Centre for Applied Statistics and Data Analytics (CAST) and School of Health Sciences (HES)

Registrations: By 13.3. through e-form:

The course is open for all Tampere University students, but limited to 30 students. If more students wish to take the course, preference is given to postgraduate students.

Target group: This course is intended for participants interested in the practical use of logistic regression (LR), with a focus on applications to epidemiology. The approach will be non-technical; some mathematics and algebra is inevitable, but only as necessary. No previous experience with LR is required, and the course should be accessible to non-statisticians with appropriate other background. Students should have some previous experience with basic statistical methods.

The course is made up of several components, as follows:

1) Lectures (two each day);

2) Practical sessions on the computer using supplied exercises and/or student’s own data (one each day);

3) Discussion of papers in the literature that have used logistic regression (this will be a group discussion during the lecture sessions).

Exercises will be provided for the practical sessions. However, students may also wish to identify their own data that could also be used during the course. Past experience has shown that participants can derive considerably greater benefit from the course if they are able to work on their own data sets. Your data might come from an ongoing research study or thesis topic, or downloaded from publicly accessible databases. Time will be available for participants to work with their own data during the practical sessions, with the assistance of the course faculty members. If your data are not yet available from a particular project, you can alternatively discuss potential future uses of the methodology with course faculty members, again during the practical sessions.

Text book: The course will be partly based on material in Applied Logistic Regression, Hosmer DW, Lemeshow S, Rodney X. Sturdivant RX, Wiley.  This is currently in a 3rd edition (ISBN: 978-0-470-58247-3, April 2013).  It is not necessary to buy this text, but if you do, it does provide many helpful details and other topics that cannot be covered during the course, because of time limitations. It therefore is a useful resource for additional reading during and after the course itself. Note that the book can be found also as an electronic version from Tampere University Library.

The timetable below is an approximate guide. Some flexibility may be required to deal with topics of interest to the group in greater depth, and we will not have time to cover not all the topics in detail (or at all).

Course notes will be distributed to students. These include copies of the lecture notes, exercises, and selections from the research literature illustrating the use of LR.

Students will be asked to read 3 research articles during the course, and they should be prepared to contribute to the group discussion of their strengths, weaknesses and interpretation. The articles will illustrate the use of LR in various epidemiologic studies. Several articles will be provided, and the group will decide which three to discuss in class.

Computer assignments will be based on the SPSS software package. Previous familiarity of students with SPSS is not required, but it would be useful. A brief orientation to data management and use of LR in SPSS will be given in the first practical session.


TIMETABLE (provisional)

Monday, March 21

Session 1: morning/1  9.00-10.30 Seminar room "Ylähylly", Building T

Overview of the course. Introductions of instructor and participants.

Introduction to examples of published papers that use logistic regression. These are for discussion on days 2, 3 and 4 of the course. Participants should read each of these articles, and they should be prepared to discuss their findings in class. You will be involved as we choose which specific papers we will discuss (we will take a vote!).

Review of basic concepts in regression in general; model formulation; how to choose the right type of regression for various types of data; review of the various purposes and objectives of regression analysis; model fitting and interpretation for linear regression.

Break: 10.30-11.00

Session 2: morning/2  11.00-12.30 Seminar room "Ylähylly"

Typical applications of LR; the logit function; simple and multiple logistic models. Data organisation and plotting the data. Parameter estimation and significance testing.

Lunch 12.30-14.00

Session 3: afternoon  14.00-15.30 computer room ML71 Arvo

Practical session: Introduction to the exercises, the computing environment and use of SPSS.


Tuesday, March 22

Session 4: morning/1  9.00-10.30 Group room RH2, Building T

Multivariable models, including predictors of various types. Partial coding method for categorical variables.  Model reduction to identify significant effects.

More on the interpretation of logistic coefficients. Binary predictor variables; odds ratio, and its approximation to relative risk. Categorisation of continuous variables. Confidence intervals for the odds ratio (OR); equivalence of LR model and contingency table approaches. Odds ratios for continuous factors. Effects of predictor variables with small units of measurement.

Break 10.30-11.00

Session 5: morning/2  11.00-12.30 Computer room ML71 Arvo

Practical session (continuing exercises and working on personal data).

Lunch 12.30-14.00

Session 6: afternoon  14.00-15.30 Group room RH2

Group discussion of literature paper I.

Coding of independent variables using referent cell (partial coding) and marginal coding methods. Polytomous (multilevel) and ordinal independent variables, and their coding.

Multivariable models; interactions; Simpson’s paradox  

Confounding in LR models.

Wednesday, March 23

Session 7: morning/1  9.00-10.30 Seminar room "Ylähylly"

Interaction and confounding in the LR model                                                                                                             

Break 10.30-11.00

Session 8: morning/2  11.00-12.30 computer room ML71 Arvo

Practical session (exercises and working on personal data).

Lunch 12.30-14.00

Session 9: afternoon  14.00-15.30 Seminar room "Ylähylly"

Group discussion of literature paper II.

Assessing model accuracy; ROC curve; overall model fit. Matched data. Sample size calculation.


Thursday, March 24

Session 10: morning/1  9.00-10.30 Seminar room "Ylähylly"

Discussion of literature paper III

Brief coverage of additional topics, as time allows. Possible topics include:

Note: the topics above are in an approximate order of priority for coverage. We will certainly not be able to discuss all of them!

Break 10.30-11.00

Session 11: morning/2  11.00-12.30 computer room ML71 Arvo or Seminar room "Ylähylly"

Because of the upcoming holiday weekend, we have decided to drop the afternoon session (which would have been Session 12) for today. For this session 11, we can choose between having a final lecture or a final practical session. There will be a vote among the students about this, to take place late on Tuesday or on Wednesday morning.

Workshop close.


Kalevantie 4, 33014 Tampereen yliopisto
puh. (03) 355 111
Maintained by:
Last update: 5.2.2016 9.52 Muokkaa