SkipCaptcha
212
edits
Changes
Jump to navigation
Jump to search
Line 95:
Line 95: + + + + + + +
Talk:PNAS Response to Letter (view source)
Revision as of 21:53, September 16, 2008
, 21:53, September 16, 2008no edit summary
::Aschlafly, you are avoiding Brossa's question. Why can't you just answer it? Brossa said that he wanted to limit the discussion to these points. The reason that he wants to do that is because it the PNAS response targets your approach to the statistical analysis. [[User:MickA|MickA]] 16:24, 16 September 2008 (EDT)
::Aschlafly, you are avoiding Brossa's question. Why can't you just answer it? Brossa said that he wanted to limit the discussion to these points. The reason that he wants to do that is because it the PNAS response targets your approach to the statistical analysis. [[User:MickA|MickA]] 16:24, 16 September 2008 (EDT)
:I don't mind addressing points one and two; I just don't think that we'll come to an agreement about them. Point one represents a misunderstanding of the figure 3, which is labeled "Alternative hypotheses for the origin of the Cit+ function..." The figure does not represent the results of the experiments and does not conflict with them. It is a cartoon of the a priori hypothesis that was generated before the experiments were performed; it is not itself the hypothesis (the map is not the country). Note that the vertical axis lacks a scale; there is no way of knowing what the actual mutation rates are ahead of time. The location of the vertical jump on the graph is abitrary; it has to lie somewhere between 0 and 31,500 generations, but that point could be anywhere. Quoting from the paper: "The historical contingency hypothesis predicts that the mutation rate to Cit+ should increase after some potentiating genetic background has evolved. Thus, Cit+ variants should re-evolve more often in the replays using clones sampled from later generations of the Ara-3 population." The hypothesis as stated does not specify a generation at which the potentiating mutation occured. The hypothesis is not that potentiation took place at generation 31,000 rather than some other generation; it is that there ''was'' a potentiating mutation ''rather than'' a rare-mutation event. The results of the experiment do not disprove the contingency hypothesis; they confirm it and suggest that the potentiating mutation took place at generation 20,000. You think that the figure is the hypothesis; I think that the hypothesis is what is stated explicitly in the text of the paper; I doubt that we'll agree.
:Point two states: "Both hypotheses propose fixed mutation rates, but the failure of mutations to increase with sample size disproves this." I disagree with this statement. The problem with comparing the 'sample sizes' in replays two and three is that the experimental conditions were similar, but not the same. One could imagine a hypothesis that men commit murder most often between the ages of 25 and 35, with samples taken from the male populations of L.A., Singapore, and London. One would find different murder rates among the men of those three cities, but still might find (or not) that murderers in those cities tend to be between 25 and 35 years old. The problem with comparing the 'sample sizes' in replays two and three is that the experimental conditions were not the same, just as Singapore is not the same as Los Angeles. The rare-mutation hypothesis does not mean that the mutation rate to Cit+ is the same for ''all experimental conditions anywhere''; just that the mutation rate is constant ''given the conditions of a particular replay''. It is possible to have different baseline mutation rates among the three replays, all of which follow the historical contingency pattern. Or the mutation rate could actually be the same across all three replays, and the results seen here are just a statistical fluke that would vanish if the replays could be run thousands of times. Either way, it's not fatal to the paper's conculsions.
:Point two also states "If the authors claim that it is inappropriate to compare for scale the Second and Third Experiments to each other and to the First Experiment, then it was also an error to treat them similarly statistically." This is also incorrect in my view. One can state that the murder rate is different between two cities, and yet the murderers have some characteristic in common. Combining results from different samples is the bread and butter of statistical analysis. Meta-analysis, for example, is used to combine the results of studies that are much different than replay experiments one, two, and three. If you wish to make a more specific argument about the '''techniques''' used to combine the three replay experiments into a single result, I'll address it. I think that the second part of point two is essentially the same as point five in that it criticizes the statistical method used to combine the results of the three replays (the Z-transform), as opposed to point three which mentions the Monte Carlo technique separately.
I therefore repeat my questions about points three and five, as stated previously. If you have other questions that I must answer first, please list them all at once, as I'm eager to move on to that discussion.--[[User:Brossa|Brossa]] 17:53, 16 September 2008 (EDT)