[R-sig-phylo] issue with cophenetic.phylo()
Thanks Klaus, The problem was that I expected logically that two nodes would have the same distance. For example with ((A,B),C), I expect d(A,C)==d(B,C). I'm sure for most people's purposes it isn't a problem that they aren't exactly equal, but if you apply functions involving the ranks of the values, then strange results can occur. This is how I came across this issue. If I use the rounded distance matrix it is fine, my functions work perfectly, but it did not occur to me to do this from the outset because I expected a matrix having particular properties. Cheers, Kelly From: Klaus Schliep [klaus.schl...@gmail.com] Sent: December 7, 2012 5:16 PM To: Kelly Burkett Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] issue with cophenetic.phylo() Hi Kelly, where is the problem? This happens if you work with floating point numbers. There is a nice link on the developer.r-project on the topic www.validlab.com/goldberg/paper.pdf The value is on my machine actually the same as .Machine$double.eps [1] 2.220446e-16 which is the smallest positive floating-point number ‘x’ such that ‘1 + x != 1’. Regards, Klaus On 12/7/12, Kelly Burkett kelly.burk...@mail.mcgill.ca wrote: Hi Everyone, I have been using the cophenetic.phylo() function and I have been having some issues with pairs of tips having distances that are different than what I would expect. The following code illustrates the issue (I have set the seed in order to reproduce these results). I generate a tree with 10 tips. There should be 9 unique non-zero distances in the matrix returned by cophenetic.phylo(). Instead there are 10 because 1.22313194 is duplicated: set.seed(1723) tree=rcoal(10) x=cophenetic.phylo(tree) y=sort(unique(as.vector(x))) y: [1] 0. 0.04371057 0.16334724 0.22292128 0.31957449 0.33656861 [7] 0.35227827 0.84410552 1.22313194 1.22313194 2.14497976 As an example, t9 should have the same distance to t8 and t2. x[t9,]: t1t3t6t9t7t2t8 t5 0.3522783 0.3365686 0.3365686 0.000 1.2231319 1.2231319 1.2231319 2.1449798 t10t4 2.1449798 2.1449798 But they are different: x[t9,t8]==x[t9,t2] [1] FALSE They are off by a very small amount: x[t9,t8]-x[t9,t2] [1] -2.220446e-16 I am posting this in case it is of interest to others. As a quick fix, I am now rounding x. Thanks, Kelly [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ -- Klaus Schliep Phylogenomics Lab at the University of Vigo, Spain ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
[R-sig-phylo] issue with cophenetic.phylo()
Hi Everyone, I have been using the cophenetic.phylo() function and I have been having some issues with pairs of tips having distances that are different than what I would expect. The following code illustrates the issue (I have set the seed in order to reproduce these results). I generate a tree with 10 tips. There should be 9 unique non-zero distances in the matrix returned by cophenetic.phylo(). Instead there are 10 because 1.22313194 is duplicated: set.seed(1723) tree=rcoal(10) x=cophenetic.phylo(tree) y=sort(unique(as.vector(x))) y: [1] 0. 0.04371057 0.16334724 0.22292128 0.31957449 0.33656861 [7] 0.35227827 0.84410552 1.22313194 1.22313194 2.14497976 As an example, t9 should have the same distance to t8 and t2. x[t9,]: t1t3t6t9t7t2t8t5 0.3522783 0.3365686 0.3365686 0.000 1.2231319 1.2231319 1.2231319 2.1449798 t10t4 2.1449798 2.1449798 But they are different: x[t9,t8]==x[t9,t2] [1] FALSE They are off by a very small amount: x[t9,t8]-x[t9,t2] [1] -2.220446e-16 I am posting this in case it is of interest to others. As a quick fix, I am now rounding x. Thanks, Kelly [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] issue with cophenetic.phylo()
Hi Kelly, where is the problem? This happens if you work with floating point numbers. There is a nice link on the developer.r-project on the topic www.validlab.com/goldberg/paper.pdf The value is on my machine actually the same as .Machine$double.eps [1] 2.220446e-16 which is the smallest positive floating-point number ‘x’ such that ‘1 + x != 1’. Regards, Klaus On 12/7/12, Kelly Burkett kelly.burk...@mail.mcgill.ca wrote: Hi Everyone, I have been using the cophenetic.phylo() function and I have been having some issues with pairs of tips having distances that are different than what I would expect. The following code illustrates the issue (I have set the seed in order to reproduce these results). I generate a tree with 10 tips. There should be 9 unique non-zero distances in the matrix returned by cophenetic.phylo(). Instead there are 10 because 1.22313194 is duplicated: set.seed(1723) tree=rcoal(10) x=cophenetic.phylo(tree) y=sort(unique(as.vector(x))) y: [1] 0. 0.04371057 0.16334724 0.22292128 0.31957449 0.33656861 [7] 0.35227827 0.84410552 1.22313194 1.22313194 2.14497976 As an example, t9 should have the same distance to t8 and t2. x[t9,]: t1t3t6t9t7t2t8 t5 0.3522783 0.3365686 0.3365686 0.000 1.2231319 1.2231319 1.2231319 2.1449798 t10t4 2.1449798 2.1449798 But they are different: x[t9,t8]==x[t9,t2] [1] FALSE They are off by a very small amount: x[t9,t8]-x[t9,t2] [1] -2.220446e-16 I am posting this in case it is of interest to others. As a quick fix, I am now rounding x. Thanks, Kelly [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ -- Klaus Schliep Phylogenomics Lab at the University of Vigo, Spain ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
