Comments for What We Have Here Is A Failure To Replicate at http://www.randi.org/site , comment 1 to 5 out of 5 comments http://www.randi.org/site Wed, 19 Jun 2013 15:49:38 +0100 FeedCreator 1.7.3 http://www.randi.org/site/index.php/swift-blog/1659-what-we-have-here-is-a-failure-to-replicate.html#comment-25023 There is an interesting analysis by Gregory Francis recently published in Psychonomic Bulletin & Review. He finds statistical evidence of publication bias in Bem's experiments, with there being only a 0.056 probability of Bem being able to achieve 8 rejections of the null hypothesis out of 9 experiments given experimental design and the effect sizes reported by Bem. Anything less than 0.1 is the standard criteria for evidence of publication bias. - Je Tue, 20 Mar 2012 18:56:48 +0100 http://www.randi.org/site/index.php/swift-blog/1659-what-we-have-here-is-a-failure-to-replicate.html#comment-25009 The fact that he peeked at the data should be enough in itself to dismiss the entire study. He cheated! Game over! - willradik Sun, 18 Mar 2012 18:32:50 +0100 http://www.randi.org/site/index.php/swift-blog/1659-what-we-have-here-is-a-failure-to-replicate.html#comment-25005 The good Dr. answers emails on a regular basis through JREF, SGU, Science Based Medicine, Neurologica and is an Assistant Professor of Neurology at Yale. I think he is pretty accessible, far more than I would want to be. If there are questions or misunderstandings with what he has written, ask them in the blog. You have stated there is an ambiguous word, but not a question. I don't think it was ambiguous, I think he meant to write words not word. Unless the typo causes the article to be unclear, is it necessary to comment? If the article is unclear, ask your question, I am sure there are others like you who may have the same question. I love his articles explaining the research. I had to take a lot of Statistics in college and I know you can make numbers dance to whatever tune you want. My first job out of school was selling copiers (exciting I know) one of the Sales managers always scolded me for not using this goofy formula for proving that owning a copier was cheaper than using Kinko's. I finally researched all the variables that I could and it was always cheaper to use Kinko's. His formula was rigged. Needless to say I didn't last long there. - xxi_centuryboy Sat, 17 Mar 2012 17:53:40 +0100 http://www.randi.org/site/index.php/swift-blog/1659-what-we-have-here-is-a-failure-to-replicate.html#comment-25004 Having been completely frustrated in my attempts to email Dr. Novella directly, I am left with only the option of posting a comment on the matter of the typos in his article. There are at least six of them, the only one of any ambiguity being this: [quote]Following that they were given two practice sessions with half of the [i][u]word[/u][/i] chosen by the computer at random. [/quote] It would be very helpful if there was a simple way to comment privately on typos and other errors in JREF articles, since the situation occurs fairly frequently. - daveg703 Sat, 17 Mar 2012 15:37:41 +0100 http://www.randi.org/site/index.php/swift-blog/1659-what-we-have-here-is-a-failure-to-replicate.html#comment-25003 Bayesian analyses, whether forward or inverse, should not be controversial. The problem I see is that most people do not understand the simple ideas involved - they seem to have been trained to use software like Statistica without knowing a thing about probabilities and they add and fiddle with parameters to get the result they want rather than doing a proper analysis of the system and coming up with a properly applied analysis. The rigorous method of analysis would typically involve creating the model before the experiment is done and if a researcher believes that the model needs some refinement afterward they need to make it clear what modifications were made and why. However, I suspect in most cases that people create their model after the fact and create models biased in favor of their desired results. - MadScientist Sat, 17 Mar 2012 14:13:37 +0100