tag:blogger.com,1999:blog-8499895524521663926.post8682878938996704849..comments2019-09-19T00:31:54.140-04:00Comments on Phylogenetic Tools for Comparative Biology: On the shape of trees with random taxa addition or subtractionLiam Revellhttp://www.blogger.com/profile/04314686830842384151noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-8499895524521663926.post-57799738208978212642013-02-28T00:43:23.110-05:002013-02-28T00:43:23.110-05:00David - Yes, this is how the function works; and I...David - Yes, this is how the function works; and I agree that this should result in trees that are not Yule-like. - LiamLiam Revellhttps://www.blogger.com/profile/04314686830842384151noreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-68866716808147361772013-02-27T02:32:20.427-05:002013-02-27T02:32:20.427-05:00Actually, looking at your code, perhaps I misinter...Actually, looking at your code, perhaps I misinterpreted your previous post about random new tips and cumsum. Now it looks like you're using it to pick an edge at random, with longer edges weighted to have a higher chance of getting chosen relative to their length? That's basically letting edges to go anywhere with uniform probability across the tree, something which would be unexpected under Yule/birth-death.David Bapsthttps://www.blogger.com/profile/17606476387441191531noreply@blogger.comtag:blogger.com,1999:blog-8499895524521663926.post-75426043625597386602013-02-27T02:06:03.659-05:002013-02-27T02:06:03.659-05:00I think the basis of what you're doing depends...I think the basis of what you're doing depends on the definition of 'at random'. In a previous post, you describe that you're adding tips as if the probability of an extra tip at any point along any edge increased with time. It isn't exactly clear, but I assume you let the probability increase linearly as branches approach the most terminal tips. This wouldn't be true under any birth/birth-death model; it's much more unlikely for a random tip to be attached further back in time than just a linearly decreasing probability. Just think of the exponential curve of a non-log-scaled LTT...<br /><br />If I remember Nee et al. rightly, the probability of an additional random tip should be the same as the probability of having a branching time of time t, as that's just the probability of having a lineage which survived for duration t.<br /><br />I think if you weighted tip addition in this way, relative to values of the birth and death rates, it would produce trees with no distortion of the estimated gamma parameter.David Bapsthttps://www.blogger.com/profile/17606476387441191531noreply@blogger.com