"Statistical tests: what it means (informally) and how to use it in neuroscience" Patricia Reynaud-Bouret (Video)

Mini-course

In neuroscience (and in many other fields), scientists are asking fundamental questions that can only be answered yes or no. For example, does the subject do the same thing twice? Do the different neurons that can be recorded during this task behave independently or not?

To answer them, scientists perform an experiment whose result is random. The hazard can be big or small. If it is too tall, nobody will be able to conclude. If it is very small, we can see the result with the naked eye. If it is moderate, then a well-defined statistical procedure is needed: the test. But the statistician performing a test cannot answer simply yes or no: it must simultaneously quantify the error committed due to the hazard. Some of these tests are very popular: chi-squared test, Student's t-test, "goodness-of-fit" test. In general, what exactly does a statistical test mean? How does he quantify the error on his answer? It is only by understanding these issues that we can avoid traditional pitfalls when performing, for example, multiple tests simultaneously (multiple tests).
All this will be illustrated on concrete situations from the analysis of the series of action potentials (spike trains): equal discharge rate test, spike train independence test, suitability tests to a model.

Video 1 Video 2

Dates

Created on March 24, 2015