PREQUENTIAL TESTS OF MODEL FIT.
SCAND J STAT
45 - 60.
A new approach to testing the goodness-of-fit of a statistical model is presented, based on its success at making successive probability forecasts for a sequence of realized data-values. This approach suggests the use of test-statistics of a certain form, whose asymptotic distribution is investigated. Heuristic arguments suggest that such a statistic will have an asymptotically standard normal distribution under the null hypothesis, under wide conditions which do not require independence. This is confirmed in specific examples, and a rigorous proof is supplied for the case of Bayesian probability forecasts in independence models.
|Title:||PREQUENTIAL TESTS OF MODEL FIT|
|Keywords:||BAYESIAN PROBABILITY FORECAST, BERNOULLI VARIABLES, GOODNESS-OF FIT, PREQUENTIAL TEST|
|UCL classification:||UCL > School of BEAMS
UCL > School of BEAMS > Faculty of Maths and Physical Sciences
Archive Staff Only