Pitfalls and paradoxes in data analysis

Jan Terje Kvaløy, Department of Mathematics and Physics (UiS) will give a talk about pitfalls and paradoxes in data analysis.

Even in very basic data analyses seemingly contradicting results can occur. For instance an effect that goes in one direction in all subgroups of the data can go in the other direction when considering all data together. A simple example of this is that it is fully possible that both the percentage of boys and the percentage of girls passing an exam increases from one year to the next, but the overall percentage of students passing the exam decreases. In this talk various pitfalls and (seemingly) paradoxes in data analysis will be discussed. Some key words are: Simpson's paradox, Lord's paradox, Berkson's paradox, immortal time bias and misinterpretation of rare events.

For information on previous and upcoming lectures in this series, see the full programme here.