tag:blogger.com,1999:blog-8499895524521663926.post8020607701358970602..comments2024-04-21T09:01:42.412-04:00Comments on Phylogenetic Tools for Comparative Biology: More robust version of brownie.lite()Liam Revellhttp://www.blogger.com/profile/04314686830842384151noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-8499895524521663926.post-55336634496894798742012-02-05T11:09:34.214-05:002012-02-05T11:09:34.214-05:00just sent you a mail with the data.
thank you ver...just sent you a mail with the data.<br /><br />thank you very much for your help!<br /><br /><br />Joan.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-32980629229284665112012-02-03T10:45:54.018-05:002012-02-03T10:45:54.018-05:00Hi Joan.
It depends what the error is, but in gen...Hi Joan.<br /><br />It depends what the error is, but in general when the rates are very small (which they will tend to be either if the scale of the traits is very small or the tree is very long) then the function, and optimization generally, will run into issues of numerical precision. That is to say, some of the values computed by the function will be smaller than the smallest value that can be stored by R.<br /><br />After multiplying the data by constant k, to get the estimated rates back on the original scale you just divide by 1/k^2.<br /><br />Another possible error could result if the rate for the lowest rate is less than 10,000 times lower than the MLE of a single, constant rate. However, this shouldn't be resolved by multiplying the data by a constant.<br /><br />If you send me your data, I would be happy to try and diagnose further.<br /><br />- LiamLiam Revellhttps://www.blogger.com/profile/04314686830842384151noreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-14887876619146758312012-02-03T10:18:08.140-05:002012-02-03T10:18:08.140-05:00Hi Liam,
I am trying to run brownie.lite on my da...Hi Liam,<br /><br />I am trying to run brownie.lite on my dataset with a tree of 150 tips and 3 character states. With the raw values of my continuous data, I never get convergence even when I increase the iteration from 2000 to 5000000, however, if I multiply by 100 my continuous variable, then it gets convergence very fast with just 2000 iterations.<br /><br />I guess there is nothing wrong in multiplying my continuous variables by a constant as this should preserve the relative values of my rates (am I right?), however why this particular detail makes converge the analysis?<br /><br />thanks a lot!<br /><br /><br />Joan.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-4923127210861984102011-07-15T16:50:28.990-04:002011-07-15T16:50:28.990-04:00Oops - the code for this is here.Oops - the code for this is <a href="http://anolis.oeb.harvard.edu/~liam/R-phylogenetics/brownie.lite/v0.4/brownie.lite.R" rel="nofollow">here</a>.Liam Revellhttps://www.blogger.com/profile/04314686830842384151noreply@blogger.com