Jean-Bernard Martens (short bio)

Eindhoven University of Technology, the Netherlands

Objectives:

This is an introductory course that doesn't require advance knowledge on statistics. The focus will be on a single specific (but frequently occurring) statistical task, i.e., comparing quantitative data collected in two distinct experimental conditions. The objective is to make clear how statistics can be used to conclude whether or not the experimental condition has an "effect" on the quantitative measure. This example is used to explain basic statistical concepts such as: histograms (that summarize the data), distributions (that model the data), matching distributions to histograms (using a quantitative criterion), confidence intervals for distribution parameters (such as averages, differences between averages, etc.) and how they can be used to make statistical inferences (i.e., draw conclusions). Participants can practice the above concepts on example data that is provided (or use their own data) with the help of a program for interactive statistics, called ILLMO.
 

Content:

Suppose we analyze current statistical practice from a user experience (UX) perspective. Most scientists interested in empirical research understand WHY statistics is relevant, although most of the motivation is extrinsic (it is needed to get quantitative research published). There however seems to be much confusion about WHAT statistics actually does (or doesn’t do), and there is virtually no understanding of HOW statistics works. This can lead to statistics anxiety, or at least confusion and uncertainty about how to interpret statistical results. 
This course explains step-by-step HOW statistics works when comparing quantitative data in two experimental conditions, and along the way explains (and lets participants practice with) the following key concepts:

1. histograms (summarizing and preparing data for statistical analysis);
2. distributions (selecting a class of models for these histograms and using the log-likelihood criteria to find the optimal distribution parameters)
3. establishing confidence intervals for model parameters (by creating log-likelihood profiles) and using these confidence intervals to make statistical inferences about significance and effect size
4. if time allows: multi-model comparison (as an alternative to hypothesis testing)

Target Audience:

Practitioners of statistics with little or no advance knowledge, or people with knowledge on traditional statistical methods (such as T-test, ANOVA, etc.) who are interested in learning an alternative approach.