Workshop 1. Use and Misuse of Significance Testing.
The originally scheduled date in January had to be cancelled -IF it gets rescheduled details will be here.
All students of social sciences learn the concept of statistical significance and techniques for testing hypotheses with sample data.
But statistical significance is not the same as substantive significance: hypothesis testing is often misapplied – surprisingly often in fact. Many research findings are reported as ‘significant’ without it being clear in what sense. Some results are significant in the statistical sense while of little or negligible value as social science. On the other hand, important effect sizes can be wrongly disregarded because they fail to pass simplistic significance tests.
This is a major issue in research across the social science from economics, to psychology. And one that is largely ignored, although there is quite a lot of literature on it.
Workshop 2. Statistical Paradoxes and Fallacies.
Wednesday February 5th - 14:00 - 17:00 - IAS Seminar Room, Milburn House
We consider Simpson’s paradox and the Regression fallacy.
Simpson’s paradox: how every component of a statistical aggregate can move in the opposite direction from the aggregate. Simpson’s paradox happens quite frequently in applied work, but is often missed. Real world examples have occurred in: education results; voting; the effect of smoking on low birth weight.
The Regression Fallacy (Galton’s fallacy). If the two variables are at different points in time, regression results can lead to highly misleading inferences about processes occurring over time. This fallacy is named after Francis Galton who detected evidence of what he called ‘regression to mediocrity’ in different human features such as stature. Another example in economics is convergence of living standards in international country cross sections. We discuss the ideas of regression lines in relation to bivariate probability distributions.
Workshop 3. Uses and abuses of the ‘Normal’ Distribution.
Wednesday February 19th - 14:00 - 17:00 - WT0.01 - Westwood Teaching Centre
Early researchers on statistical distributions thought they had found a model that could be applied to a wide range of situations in nature and society, and called it the Normal Distribution (otherwise the Gaussian distribution). The model also had nice mathematical properties making it useful as a foundation for many statistical techniques.
But there is actually nothing 'normal' about the Gaussian distribution. It limits the range of variation such that it fits many real world situations badly, notably financial data. We will consider how it became such a universal model and some of the problems this led to. We will also look at how its misapplication were a cause of the recent financial crash.
We will also consider the importance of the fundamental distinction between uncertainty and risk, look at the properties of other models including the Cauchy distribution and alternative models based on power laws.