ReproNim Statistics Module

About statistical power

Overview

Teaching: 60 min
Exercises: 120 min
Questions
  • What is power

  • Why is it important: issues with low statistical power

  • Some tools for power for neuroimaging

Objectives
  • After this lesson, you should understand what is Power, and the issues of low statistical power

Power : definition and some exercices

First, we will be doing some work on this notebook:

Power

How to work with the notebook ? There are two cases.

  1. You do not really know python, and how to install the jupyter notebook. You can still read the notebook, skipping the code sections. The notebook will introduce some definitions, and then play with different settings. But the true benefit comes if you can install the jupyter project see here and actually play with the code, for instance changing the sample size or effect size to understand better what power is.
  2. Download the notebook, and try to understand the concepts and the code. If the code is unclear, please make an issue on the repronim github site

Some simple simulations

Exercise: visit this site and set your test to be one tailed at the false positive risk of 0.05. Say a typical effect size in fMRI corresponds to a Cohen’s d of 0.5. What is the sample size that you find ? Now, relate to what we are doing in fMRI. What is the problem ? can you use this online tool to get the number of subjects for your experiment ? if yes, give a typical example, if not, explain the problem.

The Predictive Positive Value

Then, on this one:

Positive Predictive Value

Relation with confidence intervals ?

Last, you’ll need to go through :

This notebook

Key Points