tag:blogger.com,1999:blog-8499895524521663926.post2483239926075575407..comments2021-12-02T16:32:28.465-05:00Comments on Phylogenetic Tools for Comparative Biology: Testing for Pagel's λ Liam Revellhttp://www.blogger.com/profile/04314686830842384151noreply@blogger.comBlogger7125tag:blogger.com,1999:blog-8499895524521663926.post-28302908472740331172012-12-01T07:39:09.592-05:002012-12-01T07:39:09.592-05:00Hi Liam,
I ran the same analysis on another tree ...Hi Liam,<br /><br />I ran the same analysis on another tree (190 species after pruning). My results are LAMBDA = 0.9109, LR = 7.06 and p = 0.0078.<br /><br />For an upper tail one sided test we would reject the null hypothesis that Lambda = 1 for alpha = 0.05 and 0.01. We would not reject the null hypothesis that lambda = 1 for alpha = 0.001.<br /><br />Is this the right interpretation?<br />Thanks!pearsyhttps://www.blogger.com/profile/06230181517634542239noreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-85705760190773185422012-11-23T00:27:18.502-05:002012-11-23T00:27:18.502-05:00Yes - "se" is the standard error of the ...Yes - "se" is the standard error of the mean for each species. This is normally calculated for each species as the within species SD divided by the square root of the sample size.<br /><br />In the case above, we would not reject the null - but the LR was not 15 (as you say) in this particular stochastic simulation. (The LR for a P of 0.135 would be around 2.24, or so. Since this is a simulation - individual results will vary.) The rejection level for alpha=0.05 with df=1 is about 3.84.<br /><br />- LiamLiam Revellhttps://www.blogger.com/profile/04314686830842384151noreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-86860901062818854772012-11-23T00:21:23.342-05:002012-11-23T00:21:23.342-05:00Hi Alejandro. This is the default option in phylos...Hi Alejandro. This is the default option in phylosig(...,method="lambda",test=TRUE). So, for tree and data in x:<br /><br />phylosig(tree,x,method="lambda",test=TRUE)<br /><br />BTW, this indeed gives the same likelihood as the so-called "white noise" model of fitContinuous in the geiger package.<br /><br />- LiamLiam Revellhttps://www.blogger.com/profile/04314686830842384151noreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-19390900178859612582012-11-21T08:52:12.120-05:002012-11-21T08:52:12.120-05:00Hi again,
I just have a couple of short questions...Hi again,<br /><br />I just have a couple of short questions -- <br /><br />By sampling error do you mean the "se" of the value of the trait(s) of each species?<br /><br />So in the case above, we would not reject the NULL hypothesis, because LR is 15? Because p value is not that small?<br /><br />Thanks!pearsyhttps://www.blogger.com/profile/06230181517634542239noreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-48180404212828763302012-11-21T04:43:28.794-05:002012-11-21T04:43:28.794-05:00Hi Liam, to follow up on this interesting thread, ...Hi Liam, to follow up on this interesting thread, what would be your suggestion to test the other end of the spectrum, that lambda is (or not) significantly different from 0? Would a comparison with a white noise model do the trick? Or using lambdaTree() in geiger to transform the phylogeny to a star tree?<br />I enjoy the blog!<br />Cheers<br />AlejandroAlejandronoreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-48129061736650450122012-11-20T08:14:03.314-05:002012-11-20T08:14:03.314-05:00Glad to hear it. - LiamGlad to hear it. - LiamLiam Revellhttps://www.blogger.com/profile/04314686830842384151noreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-27788181278426785152012-11-18T06:37:24.101-05:002012-11-18T06:37:24.101-05:00Thanks Liam, This is very helpful!Thanks Liam, This is very helpful!pearsyhttps://www.blogger.com/profile/06230181517634542239noreply@blogger.com