ALTERNATIVE HYPOTHESES AND AIC MODEL SELECTION
Research workers in many fields are realizing the substantial limitations of
statistical tests, test statistics, arbitrary α-levels, P-values, and
dichotomous rulings concerning statistical significance. These
traditional approaches were developed at the beginning of the last century
and are being replaced by modern methods that are much more useful. These
methods rely on the concept of information loss and formal evidence. They
provide easy-to-compute quantities such at the probability of each
hypothesis/model and evidence ratios. Furthermore, simple methods allow
formal inference (e.g. prediction/forecasting) from all the models in an a
priori set (multimodel inference).
I am planning to offer several 2-day courses on the Information-Theoretic
approaches to statistical inference during February-June, 2014. These
courses focus on the practical application of these new methods and are
based on Kullback-Leibler information and Akaikes information criterion
(AIC). The material follows the recent textbook,
Anderson, D. R. 2008. Model based inference in the life sciences: a
primer on evidence. Springer, New York, NY. 184pp.
A copy of this book, a reference sheet, and several handouts are included in
the registration fee. These courses stress science and science philosophy
as much as statistical methods. The focus is on quantification and
qualification of formal evidence concerning alternative science hypotheses.
These courses can be hosted, organized, and delivered at your university,
agency, institute, or training center. I have given >60 of these courses
and they have been well received. The courses are informal and discussion
and debate are encouraged. Further insights can be found at
informationtheoryworkshop.com
If you are interested in hosting a course at your location, please contact
me.
Thank you.
David R. Anderson
[email protected]
