tag:blogger.com,1999:blog-8499895524521663926.post6982253980330948512..comments2021-09-25T14:09:23.073-04:00Comments on Phylogenetic Tools for Comparative Biology: New version of phylosig with control of multidimensional optimization migrated to the userLiam Revellhttp://www.blogger.com/profile/04314686830842384151noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-8499895524521663926.post-88257578759804063732012-02-11T11:11:06.222-05:002012-02-11T11:11:06.222-05:00Oh, and in addition the updates reject values of s...Oh, and in addition the updates reject values of start that don't satisfy this criteria.Liam Revellhttps://www.blogger.com/profile/04314686830842384151noreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-77688069693499007622012-02-11T11:10:37.708-05:002012-02-11T11:10:37.708-05:00Hi Matt.
Issue #1 is indeed an error.
Issue #2 i...Hi Matt.<br /><br />Issue #1 is indeed an error.<br /><br />Issue #2 is that start should be a vector containing the starting values for ML optimization of sigma^2 and lambda. Thus, lambda should probably be on the interval of 0 to 1 (or the maximum value of lambda). And sigma^2 should be on the interval 0 to Inf, not including 0.<br /><br />I have posted an updated version.<br /><br />- LiamLiam Revellhttps://www.blogger.com/profile/04314686830842384151noreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-6754093538002986432012-02-10T22:15:39.862-05:002012-02-10T22:15:39.862-05:00Thanks so much Liam! I have two comments/questions...Thanks so much Liam! I have two comments/questions:<br /><br />1) There appears to be a typo on line 126-- when test=T and me=T, convergence is spelled wrong so it always returns NULL when testing against lambda=0.<br /><br />2) What are the kinds of values I should be using for "start"? I noticed that I needed to provide two numbers. The first number appears to be related to a random scaling of the squared mean of PIC for the trait on the tree, and the second is a random number between 0 and 1. <br /><br />On my real data I would often get the "matrix is singular" error, but not every time I ran it! Certain combinations of the two values would work for some traits, while others would not. <br /><br />Do you have any suggestions for improving these starting values so I can be assured that a trait does not run into this intermittent problem?Matthttps://www.blogger.com/profile/17746570756439422280noreply@blogger.com