tag:blogger.com,1999:blog-3558941018059040035.post6846333851209330114..comments2017-10-22T08:26:41.859-07:00Comments on The pathological science: Psychology, skepticism, and statistics: For a Type 1 error rate of 5%, accept H1 if BF10 > 1 (wait, what?)Matthttp://www.blogger.com/profile/15143483413289978878noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-3558941018059040035.post-4763328694695465242016-11-14T13:48:18.602-08:002016-11-14T13:48:18.602-08:00Thanks Alex! The posts you link to definitely do a...Thanks Alex! The posts you link to definitely do a much more solid and comprehensive job at looking at this than I have - my post is really more just pointing out the curious way the error rates pan out in the scenario Daniel discussed. I wasn't aware of those posts so thanks (and I've added your RSS feed now :) )Matthttps://www.blogger.com/profile/15143483413289978878noreply@blogger.comtag:blogger.com,1999:blog-3558941018059040035.post-3456262098191884622016-11-14T13:20:35.070-08:002016-11-14T13:20:35.070-08:00You might be interested in these other posts that ...You might be interested in these other posts that followed up Daniel's. Tim did similar simulations, and I showed how you can easily do this stuff analytically by using the t CDF (since t is a sufficient statistic in this special case of the BF. Not always!).<br /><br />https://timvanderzee.wordpress.com/2016/07/19/error-control-p-values-versus-bayes-factors/<br /><br />https://alexanderetz.com/2016/07/20/a-quick-comment-on-recent-bf-vs-p-value-error-control-blog-posts/<br /><br /><br />I would also note that simulating from a single alternative value is not really the right simulation for this model specification. The proper Bayesian simulation would sample from the alternative prior in such a way as to obtain the prior predictive distribution. Such as is done in Jeff's paper here:<br /><br />http://link.springer.com/article/10.3758/s13423-014-0595-4/fulltext.htmlAlexander Etzhttps://www.blogger.com/profile/07682503927955600974noreply@blogger.com