On Oct 22, 3:49 am, djh <halitsk...@att.net> wrote: > [Explanatory note to Jacques/Arthur/Marvin: the "e" that Ray is > talking about here is an average O-F value computed from the O-F > values of the individual base-pair doublets in a set of one or more > codons, the way we always have.] > > You wrote: > > As usual, I'm still several steps behind you. I'm worried about e. > > It may be on a 0...max scale that is easily interpretable and that > > relates (approximately, on average) linearly to temperature, but that > > doesn't mean that it's legitimate to use e in a ratio or take its log > > -- both of which are things that should not be done with Celsius or > > Fahrenteit temps, but can be done with Kelvins. In other words, I > > fear the zero on the e scale is a Celsius zero, not a Kelvin zero. > > If I were not completely automated at this point, I would of course be > concerned with the time it would take to re-run the entire analysis > using a new definition of our "e" based on an O-F index that has > somehow been converted to avoid the problem you note above. > > But since a complete re-run is now only a matter of a day and a half, > my concern is solely theoretical. > > Namely, is it possible to "convert" the O-F scale so as to give us a > defensible basis for our computation of "e" as an average across a set > of dicodons?
It can probably be done, but I still don't understand e completely. All the individual values are in [0, 2.86]. If they are averaged and then divided by the maximum possible (i.e., 2.86), how do values > 1 arise?
> > Please let me know your thinking here and I will of course pass on > your thoughts to Jacques and Arthur. > > Thanks as always.