Information Technology Reference
In-Depth Information
Chapter 22
Paper Rejected (p>0.05):
An Introduction to the Debate
on Appropriateness of Null-
Hypothesis Testing
Mark. D. Dunlop
University of Strathclyde, UK
Mark Baillie
University of Strathclyde, UK
ABSTRACT
Null-hypothesis statistical testing has been seriously criticised in other domains, to the extent of some
advocating a complete ban on publishing p-values. This short position paper aims to introduce the argu-
ment to the mobile-HCI research community, who make extensive use of the controversial testing methods.
INTRODUCTION
medicine (e.g. (Ioannidis, 2005)), education (e.g.
(Cliner, Leech, & Morgan, 2002)), political science
(e.g. (Gill, 1999)) and other branches of computer
science (e.g. (Demsar, 2008)) for example, it has
yet to take root in HCI, despite our inheritance
of many methods from experimental psychology.
Nor has it had much effect on text books (Cliner,
Leech, & Morgan, 2002) from which we and our
students typically learn our statistics - a serious
problem because many researchers have to teach
themselves statistics and one that is compounded
because many of the expectations for good practice
The approach of statistical analysis using a null-hy-
pothesis testing has been heavily criticised in other
domains. Reinvigorated by Cohen and Meehl's
seminal papers (Cohen, 1994) (Meehl, 1990) there
has been a long running debate in experimental
psychology that has led to The American Psycho-
logical Association considering, but not going so
far as, a complete ban on reporting of p-values
(Wilkinson, 1999). While this debate has reached
Search WWH ::




Custom Search