The 'small n problem' in sports research.

ACEMS / SportsStats2017

A place for discussing ideas, and following up collaborations, for the March 2016 Sport Statistics Workshops at the ACEMS - QUT node.

1 stars 1 forks source link

Open jesse-jesse opened 7 years ago

cdrovandi commented 7 years ago

Bayesian methods are quite useful here for 2 reasons:

prior information from past studies or experts can be incorporated
Bayesian methods lead to shrinkage in regression coefficients, avoiding overfitting when the number of predictors is greater than the number of observations.