12th February 2015 - University of Stirling

Research Seminar Programme
12th February 2015
Professor Roger Watt
University of Stirling
Host: Professor Ronan O’Carroll
Statistical Journeys with BrawStats
The main purpose of this talk is to explain to colleagues the type of statistical analysis/argument that they may see in dissertations this year
and how it is produced. I will be emphasising the issue of low numbers of participants.
However, I will set this in the framework of a number of statistical journeys where the material is allowed to develop by asking a question and
then following the answer and its implications. For example, one can ask the question what is the distribution of expected p-values in a simple
correlation analysis? The answer is that the distribution can be very wide in typical circumstances with luck playing a large role in whether a
reportable result is obtained. This leads to the question what could I do to manage this? Exploring the consequences of sample size for the
distribution of expected p-values produces some possibilities, but of course, an optimal sample size (which you’d like to know) depends on the
population correlation (which you don’t know). One can see easily how the temptation to keep on increasing sample size will arise and how unsafe it is to do so. This journey then takes a left turn.
My opinion is that these journeys graphically illustrate the current operational problems that are associated with traditional statistics. I will of
course allude to the so-called new statistics and explain how they differ from what we are used to and how they may be better.
I will also sketch out what I see as a potentially interesting and useful educational statistics journey for teaching the subject to students.
I will place the whole talk in the context of the journey that the development of statistics has followed over the last 150 years. The reason for
this is that much of the present state of affairs is a consequence of the historical journey – I will argue that “history is bunk” and that there are
benefits to stripping away the relics of heroic pre-computer methods of data analysis.
Seminars are free of charge and held on
Thursdays @ 4:00 pm Room 3A94
Refreshments are served upon arrival * All are welcome
Psychology — School of Natural Sciences — University of Stirling — Stirling FK9 4LA
Tel: 01786 467640 - Email: [email protected]
Web: http://www.stir.ac.uk/natural-sciences/about-us/psychology/