tag:blogger.com,1999:blog-6894866515532737257.post3958504272475252593..comments2024-03-27T01:01:09.785-07:00Comments on Probably Overthinking It: Statistical inference is only mostly wrongAllen Downeyhttp://www.blogger.com/profile/01633071333405221858noreply@blogger.comBlogger5125tag:blogger.com,1999:blog-6894866515532737257.post-55019520710383177792015-03-11T13:28:16.418-07:002015-03-11T13:28:16.418-07:00Allen, his/her point is that the p value is "...Allen, his/her point is that the p value is "the probability of the data given chance". It is not "the probability of chance given the data".dustin lockehttps://www.blogger.com/profile/12240156576005704547noreply@blogger.comtag:blogger.com,1999:blog-6894866515532737257.post-56179634399275280132015-03-10T05:11:52.847-07:002015-03-10T05:11:52.847-07:00I'm not sure I understand your objection. I w...I'm not sure I understand your objection. I was using "due to chance" as a shorthand for "under the null hypothesis", since the null hypothesis is a model of random variation if there is no actual effect.<br /><br />The sentence you quoted is one of four possible explanations for an apparent effect: it might be caused by random variation in the absence of a real effect.<br /><br />As you said, the p-value is the probability of the apparent effect under the null hypothesis, which is the probability of the effect under (at least a model of) random chance.<br /><br />Can you clarify what you are objecting to?Allen Downeyhttps://www.blogger.com/profile/01633071333405221858noreply@blogger.comtag:blogger.com,1999:blog-6894866515532737257.post-59980598640532965072015-03-09T21:08:36.570-07:002015-03-09T21:08:36.570-07:00GamingLifer nails it. Allen, you appear to have m...GamingLifer nails it. Allen, you appear to have made *the* mistake the editors are so concerned about re p-values.dustin lockehttps://www.blogger.com/profile/12240156576005704547noreply@blogger.comtag:blogger.com,1999:blog-6894866515532737257.post-70635172920251218192015-03-08T18:21:51.581-07:002015-03-08T18:21:51.581-07:00"The apparent effect might be due to chance; ..."The apparent effect might be due to chance; that is, the difference might appear in a random sample, but not in the general population."<br /><br />p-values don't tell you this, either. All a p-value tells you is the probability of your data (or data more extreme) given the null is true. They say nothing about whether your data is "due to chance."GamingLiferhttps://www.blogger.com/profile/02195336931104729563noreply@blogger.comtag:blogger.com,1999:blog-6894866515532737257.post-68901014915615703422015-03-02T15:40:40.361-08:002015-03-02T15:40:40.361-08:00You're still giving undue credit to CIs as mea...You're still giving undue credit to CIs as measures of precision: http://www.ejwagenmakers.com/submitted/fundamentalError.pdf and I think your recommended approach - essentially sticking with orthodox statistical inference - is unjustifiable. We know that it's fundamentally, conceptually flawed, fosters bad science, lends itself to misuse and misinterpretation even by experts, and that there is a well-founded, good science-fostering, reason and intuition-congruent alternative. It just doesn't make sense to recommend sticking with it.Paul Hayeshttps://www.blogger.com/profile/04309125585593320043noreply@blogger.com