tag:blogger.com,1999:blog-987850932434001559.comments2017-02-24T02:45:43.789-08:00The 20% StatisticianDaniel Lakensnoreply@blogger.comBlogger712125tag:blogger.com,1999:blog-987850932434001559.post-51568226669919092792017-02-24T02:45:43.789-08:002017-02-24T02:45:43.789-08:00Thank you!!!
EnricoThank you!!!<br /><br />EnricoEnrico Glereanhttp://www.blogger.com/profile/16674832915668714617noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-33980430465458071442017-02-24T02:33:28.078-08:002017-02-24T02:33:28.078-08:00Yes, it should be possible - either take the 90% C...Yes, it should be possible - either take the 90% CI approach, or use dedicated software: https://ncss-wpengine.netdna-ssl.com/wp-content/themes/ncss/pdf/Procedures/NCSS/Testing_Equivalence_with_Two_Independent_Samples.pdf - might program it into the package in the future!Daniel Lakenshttp://www.blogger.com/profile/18143834258497875354noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-54827849002343792552017-02-24T01:54:17.588-08:002017-02-24T01:54:17.588-08:00Thank you for this! It has been very useful in a p...Thank you for this! It has been very useful in a paper we are finalising. Quick question: is there a TOST equivalent for Likert-type data (e.g. sign rank test instead of t-test)? Would it be enough to convert likert scores to ranks?<br /><br />Here something similar I have found: http://stats.stackexchange.com/questions/52897/equivalence-tests-for-non-normal-data<br /><br />Enrico Glerean, www.glerean.comUnknownhttp://www.blogger.com/profile/16674832915668714617noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-77426187198278461582017-02-21T09:46:05.369-08:002017-02-21T09:46:05.369-08:00I sat in my lounge room in Australia watching a do...I sat in my lounge room in Australia watching a documentary that was telling the story of a young Rohingya girl. https://www.fiverr.com/sabakhan695<br />john smithhttp://www.blogger.com/profile/07655303417027257101noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-73946100180359827062017-02-20T03:49:03.157-08:002017-02-20T03:49:03.157-08:00Really nice blog shared. Keep sharing more updates...Really nice blog shared. Keep sharing more updates with us.<br />plots near tcs indorehttp://www.blfbhumi.com/about-us/noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-75588939157686237212017-02-16T13:25:37.173-08:002017-02-16T13:25:37.173-08:00Thanks for this interesting blog post, Daniel. I&#...Thanks for this interesting blog post, Daniel. I've created a follow-up that shows cases in which TOST+NHST yield conflicting decisions, which can never happen with the HDI+ROPE procedure. It's here: http://doingbayesiandataanalysis.blogspot.com/2017/02/equivalence-testing-two-one-sided-test.htmlJohn K. Kruschkehttp://www.blogger.com/profile/17323153789716653784noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-24363030584744124132017-02-16T13:24:49.379-08:002017-02-16T13:24:49.379-08:00This comment has been removed by the author.John K. Kruschkehttp://www.blogger.com/profile/17323153789716653784noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-30531319902663292017-02-14T01:00:57.852-08:002017-02-14T01:00:57.852-08:00Hi, that looks like it will be much easier to use ...Hi, that looks like it will be much easier to use in the future! Excellent!Daniel Lakenshttp://www.blogger.com/profile/18143834258497875354noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-898102092936956462017-02-13T17:58:13.756-08:002017-02-13T17:58:13.756-08:00agreed...
It is important to emphasize that in o...agreed... <br /><br />It is important to emphasize that in one instance you have a measurement of belief and the other you can only make a yes or no decision that may or may not update your belief but in the end provides no measurement of that belief. That is not a trivial pedantic distinction to ignore. People end up coming away thinking that the frequentist method provides a measure that it does not.Unknownhttp://www.blogger.com/profile/00227235335343168838noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-38609932056081186822017-02-13T17:40:20.953-08:002017-02-13T17:40:20.953-08:00Hi Daniel, Glad you find BEST useful. If you just ...Hi Daniel, Glad you find BEST useful. If you just want an HDI, use HDInterval::hdi; same as BEST::hdi but faster for large objects and you don't need to install JAGS. The next version of BEST will 'Depend' on HDInterval.Michael Meredithhttp://www.blogger.com/profile/09509778781847126034noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-22663138064826088582017-02-13T07:56:34.792-08:002017-02-13T07:56:34.792-08:00Hi Heather - you can use it for equivalence with a...Hi Heather - you can use it for equivalence with a non-zero range as well. For example, using the TOST for one-sample, you could test whether a score is equivalent to guessing average (e.g., 0.5). <br /><br />I'm not sure this is addressed in the Coursera course - but I will be updating the equivalence assignment in the future now my own paper on this is out, and will add a non-zero example!Daniel Lakenshttp://www.blogger.com/profile/18143834258497875354noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-7961250058665800192017-02-13T05:11:31.206-08:002017-02-13T05:11:31.206-08:00Nice post, Daniel. Thank you. Does TOST or ROPE re...Nice post, Daniel. Thank you. Does TOST or ROPE require including 0 in the interval, or can one or both be used to examine equivalence within nonzero ranges? (It's possible your Coursera course addressed this for TOST. If so, color me embarrassed that I can't remember.)Heather Urryhttp://www.blogger.com/profile/13691556381219306479noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-78097488060212515752017-02-12T13:02:51.652-08:002017-02-12T13:02:51.652-08:00So I would argue that a region of practical equiva...So I would argue that a region of practical equivalence (ROPE) is both computationally and conceptually very different from a equivalence testing. <br /><br />A ROPE is a very simple concept, it's still a good concept, but it's also very simple. It's a range of differences between any two parameters where, if the "underlying" difference falls in that range, it isn't large enough to be of interest. This is a very general concept not tied to a specific model or specific parameters. You could use it for differences in means, scale parameters, or any other exotic parameters. You could use it for simple group models, or for more advanced models where it's not even clear how you would calculate a p-value. Even if it was originally introduced together with BEST you can use it with *any* Bayesian model, and once you have fitted a Bayesian model it's straight forward to calculate how much probability is in or out of the ROPE (or use an HDI if you want to).<br /><br />Equivalence testing is something different, it's a procedure that requires you to use a model and a parameter where you can calculate p-values. Using a ROPE can be seen as a way of summarizing a posterior distribution, while equivalence testing relies on p-values. And I would say that there is a big conceptual difference between posterior probabilities and p-values even if they, in a few select cases, are numerically similar.Rasmus Bååthhttp://www.blogger.com/profile/16575386339856902265noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-44297582581243986792017-02-07T05:02:19.214-08:002017-02-07T05:02:19.214-08:00Quantitative data depicts the quality and can be s...Quantitative data depicts the quality and can be scrutinized, but measuring it precisely is daunting enough; in contrast quantitative data can be easily measured and is depicted in number or amount. <a href="http://www.statisticaldataanalysis.net/comparative-data-analysis-definition/" rel="nofollow">comparative data analysis</a><br />Robert Lindehttp://www.blogger.com/profile/04258403403329675528noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-46068633328221463422017-02-07T05:01:18.252-08:002017-02-07T05:01:18.252-08:00This comment has been removed by the author.Robert Lindehttp://www.blogger.com/profile/04258403403329675528noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-38357978434867075402017-02-03T09:17:48.213-08:002017-02-03T09:17:48.213-08:00I finally got convinced there's no problem wit...I finally got convinced there's no problem with doing TOST and t-test on the same set of data. What is still unintuitive to me is that this combination makes significant results more frequent without inflating alpha. I understand the reason is that both null hypotheses are mutually exclusive. Thanks for your patience :-) matanhttp://twitter.com/mazormatannoreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-62326494425029104032017-02-02T02:43:20.232-08:002017-02-02T02:43:20.232-08:00My full response here: https://medium.com/@mazorma...My full response here: https://medium.com/@mazormatan/cant-have-your-tost-and-eat-it-too-f55efff0c85e#.a2vl4umpqmatanhttp://twitter.com/mazormatannoreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-83559159126723583832017-02-01T05:39:16.649-08:002017-02-01T05:39:16.649-08:00Thank you for this valuable information, it is rea...Thank you for this valuable information, it is really useful. Miguel Landa Blancohttp://www.blogger.com/profile/01204900184465837271noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-49602850287906187342017-02-01T00:36:10.982-08:002017-02-01T00:36:10.982-08:00Thanks for the clarification, Bill!
"The Bay...Thanks for the clarification, Bill!<br /><br />"The Bayesian prior was a somewhat subjective illustration of how someone who believed in the effect would describe that belief."<br /><br />I appreciate that it's difficult to quantify subject beliefs, particularly if you don't share them. But would someone holding this belief expect that the ratio of the treatment effect and the standard deviation was the same for all ten outcome variables? I.e., that while the variability of the data for DV1 may be larger than that for DV2, the treatment effect for DV1 would be correspondingly larger as well to produce the same ratio?<br /><br />This isn't a criticism of your study. But I don't understand, in general, why one would express predictions in standardised effect sizes. 'Because we don't know about the raw effect size' isn't really a strong argument, because you need the raw effect size to calculate the standardised effect size.<br /><br />@Daniël: Thanks for the link.Janhttp://www.blogger.com/profile/17765078332699225416noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-81726970004303223132017-01-31T07:40:10.070-08:002017-01-31T07:40:10.070-08:00I don't understand. In both cases the error ra...I don't understand. In both cases the error rate of each of the tests is kept under alpha, and in both cases it's meaningless to talk about alpha for single tests because the tests are the same, only with different rejection areas. The serious problem here is that at least one of the null hypotheses is always false: either d!=0, or d==0, and then it lies within the equivalence interval. This makes alpha completely meaningless, because there is no null distribution (so to correct my previous comment, alpha is not 1, it is just undefined).matanhttp://twitter.com/mazormatannoreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-44616492479332902342017-01-31T06:07:52.746-08:002017-01-31T06:07:52.746-08:00I think there is a difference with the 2-tail exam...I think there is a difference with the 2-tail example, namely that in that case, you are testing: 1) d > 0, 2) d < 0, 3) d = 0. If d = 0 is true, but you test 1 and 2 with 5% alpha, your overall alpha is actually 10% when d = 0. But with equivalence tests, the t-test has a 5% error rate when d = 0, and the TOST test only has a 5% error rate when d <> 0. I think that's a difference.Daniel Lakenshttp://www.blogger.com/profile/18143834258497875354noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-45024749701680506252017-01-31T05:33:53.148-08:002017-01-31T05:33:53.148-08:00Well, you can use the exact same argument to justi...Well, you can use the exact same argument to justify doing a right tailed t test and move on to left tailed t test only if not significant. Either you make type 1 error on the first test or on the second, can never be both. In the example from my previous comment, you will reject at least one hypothesis for every possible combination of x̄ and σ, so your alpha is 1! Each of the tests is legitimate, its the combination that's problematic.matanhttp://twitter.com/mazormatannoreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-17305175226856599282017-01-31T05:19:47.969-08:002017-01-31T05:19:47.969-08:00Hi, yes, they are performed on the same data, but ...Hi, yes, they are performed on the same data, but either 1) the true effect is 0, so you can make a Type 1 error for the t-test, but not for the equivalence test, or 2) The true effect is <> 0, so you can make a Type 1 error for the equivalence test, but not for the t-test. Doesn't this solve the problem? It's an interesting question, and it is very well possible that I am missing something.Daniel Lakenshttp://www.blogger.com/profile/18143834258497875354noreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-27317266854973535032017-01-31T04:49:44.552-08:002017-01-31T04:49:44.552-08:00Corrections: should be:
Rejection areas for the fi...Corrections: should be:<br />Rejection areas for the first test are t>1.98 OR t<1.98, i.e., x̄/(σ/sqrt(100))=10*x̄/σ>1.98 OR 10*x̄/σ<-1.98, i.e., x̄/σ>0.198 OR x̄/σ<-0.198<br /><br />more importantly, rejection areas for the second test are 1.66/sqrt(100)-0.5<x̄/σ<-1.66/sqrt(100)-0.5, i.e., -0.33<x̄/σ<0.33<br />The important point holds though - both have the same null distribution given the df, and thus should not be treated as independent tests.matanhttp://twitter.com/mazormatannoreply@blogger.comtag:blogger.com,1999:blog-987850932434001559.post-89682593213439222822017-01-31T04:47:31.662-08:002017-01-31T04:47:31.662-08:00This comment has been removed by the author.Tan Zor Δhttp://www.blogger.com/profile/15780600159750105876noreply@blogger.com