2013 FLAMES SUMMER SCHOOL in Methodology and Statistics


Date: from Monday 16 September 2013 to Friday 20 September 2013
Location:
Leuven, Belgium

Link: http://www.flames-statistics.eu/

Attending Members



Description

Flanders Training Network for Methodology and Statistics (FLAMES) invites you to a five-day Summer School aimed at providing practical methodological and statistical training to young researchers (particularly PhD students) from all disciplines.

Module II: Basic Statistical Principles and Beyond

The module revisits the foundations for statistical data analysis. The module consists of the general principles starting from a descriptive analysis to hypothesis testing. The framework is first approached from a normality perspective leading to the methodology established in standard Bachelor courses. Thereon, bells and whistles are designed for dedicated modifications to the framework under specific measurement problems. The classical framework is extended towards non-Gaussian hypothesis testing and testing under heteroscedastic conditions. On top of that, we deal with the problem of missingness. The framework and its customizations under measurement problems are illustrated on real-life data.

Topics:

  1. Descriptive statistics
  2. Graphical exploratory statistics
  3. Hypothesis testing and ANOVA
  4. Post-hoc analysis and multiple comparisons
  5. Missingness
  6. Non-parametric tests

The module is taught in 4 sessions of 1.5 hours where principles will be introduced, and 4 sessions of 1.5 hours with computer exercises (software: SPSS or SAS, to be chosen by the participant).

 

Module III: Flexible Regression for Practical Research

The linear regression model is a powerful tool that forms the basis for analyzing a wide variety of data, yielding estimates, predictions and hypothesis tests. Most basic linear regression courses model simple straight lines and end with only a basic verification of the model assumptions. However, unlike with textbook examples, translating real-life research questions into this flexible framework is not always easy and demands skill, sometimes creativity, and the ability to recognize when the limits of the model are reached and more advanced methodology is needed. This course starts where basic courses end, investigating typical set-ups through a number of real-life cases, showcasing the potential of good robust use of the linear model, creative adaptations and more advanced modeling options. Through case-studies and published results, we focus on the practical interpretation, and on ways to make the mathematics work for you rather than the other way around. Hands-on practical sessions will be organized separately for users of different software packages (R, SAS and SPSS - if registration permits).

During the sessions, we will

  1. revisit the basic model, recapping the assumptions and stressing their practical importance;
  2. discuss the inclusion of multiple predictors, categorical covariates, interactions and higher order terms;
  3. illustrate how the choice of parameterization can be used to your advantage;
  4. see how specific research questions can be optimally addressed using variations on the straight regression line using, for example, piecewise regression models;
  5. investigate non-linear trends in the data using flexible, non-parametric methods such as kernel smoothing and splines;
  6. give examples of the strange behavior of real-life data (e.g. outliers, non-constant variances, data artifacts, dependence,...), and how to cope with this in practical analysis;
  7. finally give tips on how to focus on improving prediction and automatically select better models.

The module will consist of 4 lectures focusing on the data structure, the model and its interpretation through real-life case-studies, and 4 hands-on computer labs (parallel sessions per software package).

©Copyright 2017. All rights reserved. Sitemap Read our disclaimer