![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Entry tags:
EUfuckingREKA
WAIT, I just wrote in a note to myself so I wouldn't forget the BLINDING INSIGHT I'd just had:
You were talking about looking at the differences between observed and expected usage frequency, right, and doing the graph in terms of the logarithms of these figures rather than the straight figures themselves, because with the straight figures the differences were obscured by the sheer scale of the thing. Only the trouble was, and you were planning to talk about this in a paragraph loosely describable as "well, but it's only a master's thesis", when you subtract logarithms you're actually dividing "real" numbers, meaning that thing you keep on describing as a difference is actually a ratio. Not that there's anything wrong with this (because, see above, only a master's thesis), but, still, it's not really quite exactly doing what you set out to do.
Only, listen, listen, instead of graphing logObs-logExp, is there any good reason you can't graph log(Obs-Exp), which will really honestly be a log of the difference?
[trembles with the grooviness of this, and the fact that it occurred to self rather than to supervisor]
I mean, I still have to do the bloody analysis. But, but, but. Yay?
You were talking about looking at the differences between observed and expected usage frequency, right, and doing the graph in terms of the logarithms of these figures rather than the straight figures themselves, because with the straight figures the differences were obscured by the sheer scale of the thing. Only the trouble was, and you were planning to talk about this in a paragraph loosely describable as "well, but it's only a master's thesis", when you subtract logarithms you're actually dividing "real" numbers, meaning that thing you keep on describing as a difference is actually a ratio. Not that there's anything wrong with this (because, see above, only a master's thesis), but, still, it's not really quite exactly doing what you set out to do.
Only, listen, listen, instead of graphing logObs-logExp, is there any good reason you can't graph log(Obs-Exp), which will really honestly be a log of the difference?
[trembles with the grooviness of this, and the fact that it occurred to self rather than to supervisor]
I mean, I still have to do the bloody analysis. But, but, but. Yay?