Regression Models as a Tool in Medical Research (2011)

Aim:

The participants should become familiar with the basic concepts and techniques in using regression models in medical research. They should be enabled to perform analyses of their own data, and to interpret, communicate and publish the results. They should also understand the basic potentials and limitations in using regression models and get some inspiration for a more effective and a more understandable use of regression models.

Target group:

Postgraduate students and young researchers from the health sciences and related fields who want to work with regression models in their own research or wish to understand regression model based analyses found in the literature.

Topics covered:

• Why using regression models? The basic interpretation and use of regression coefficients
• The basic output from regression analyses: Effect estimates, confidence intervals and p-values
• The basic regression models: Classical regression, logistic regression, Cox regression
• Presenting the results of a regression analysis: Meaningful tables and graphs
• Adjusted effects: Using regression models in the presence of confounding
• Categorical and ordinal covariates: How to incorporate in a regression analysis
• Common pitfalls in using regression models: Causality vs. association, relevance vs. significance, statistical models vs. real world models
• Variable selection: Which variables should be in my model?
• Effect modification and interactions: Assessment and interpretation
• Non-linear effects: Some techniques and their meaningful use
• Risk scores and predictions: Construction and assessment of precision
• Incomplete data: Selecting a strategy to handle missing values
• Measurement error: The impact on a regression analyses
• Goodness of a model: How good is a model and how good need it to be?
• Working with regression coefficients: How we can improve and support their interpretation
• Modern techniques: Robust inference, clustered data, longitudinal data
• Sample size and sample selection: How many subjects from which population?
• Alternatives to regression models: Stratification, trees, correlation, standardization and propensity score

All topics will be illustrated by examples from various fields of medical research like prognostic studies, epidemiological risk factor studies, experimental studies, diagnostic studies and observational studies.

Prerequisites:

Only some basic knowledge of statistical terms like mean, p-value and confidence interval is required. The participants should have access to the Internet.

Registration fee:

800 EURO (inluding a one year license of Stata)
750 EURO (without Stata license)

Registration:

Starting December 8, 2010, you can register for the course at http://www.isqr.intercongress.de. You can pay directly with a credit card, or you can transfer your registration fee to a bank account.

Practicalities:

The course is offered as a web based distance learning course over a period of 16 weeks. The participants are expected to download and read about 25-35 pages of teaching material (pdf-files) each week. Examples from the teaching material can be found here. The teaching material covers an introduction into the concepts and methods as well as instructions how to apply the methods using the statistical package Stata. Each week there will be practical exercises, with solutions provided one week later. There will be a discussion list allowing communication among the participants and with the teaching staff. Any question to the teaching staff will be answered within 24 hours.

Use of Stata:

A one year license of Stata 11 is included in the registration fee. Stata is available for Windows, Unix and Mac. For further details visit the Stata home page.
Participants unfamiliar with Stata will get a short introduction prior to the course. The examples and exercises of the course require access to Stata 10 or Stata 11. The Intercooled version of Stata is sufficient.

Course certificate:

A course certificate is given to all participants, who participated in a multiple choice test and delivered a final written exercise.

Course credits:

8 ECTS

Stipends:

A few stipends are available for participants who cannot afford the registration fee. For further information contact stipend@isqr.uni-freiburg.de

Teaching staff:

Werner Vach and Primrose Beryl, Clinical Epidemiology, Institute of Medical Biometry and Medical Informatics, University of Freiburg

Course administration:

Per Berg and Monika Richards, International School of Quantitative Research, University of Freiburg
email: info@isqr.uni-freiburg.de

Important dates:

Course schedule:

Each teaching week starts on a Thursday. The material for each week will be available already one the preceding Tuesday. The discussion of the contents of each teaching week on the discussion list should be limited to the teaching week and the following two weeks.

Evaluation of previous courses:

The evaluation results of a of this course can be found here.