keith at eve.cchem.berkeley.edu
Sat Dec 5 23:02:53 EST 1998
In <7opiGDf98QgB-pn2-dy9Z1ZHxmmZ4 at rnaworld.bio.ukans.edu> PGegen at UKans.nolospamare.edu (Dr. Peter Gegenheimer) writes:
>Yes -- it IS _always_ most accurate to fit the untransformed data to the equation
>which describes the relationship between X and Y; you are guaranteed to get the best
>estimate for the kinetic constants. Not only for the reasons Simon mentions, but also
>because when you transform the variables, you also transform their experimental error
>so that it no longer has a ~Gaussian distribution. Further, the transformations in
>which one axis contains both V and S violate the assumtions of regression analysis,
>in that the X and Y axis variables are no longer separate (you no longer have an
>independent and a dependent variable), and (so certain transforms) the experimental
>error is no longer exclusively in the Y axis.
While I generally agree with all of these points, I have to say
that I still think there's a place for the linear transform plots.
If your data has some kind of systematic deviation from
Michaelis-Menten kinetics, it's often much easier to recognize
that on a linear transform than on the non-linear plot.
So I generally do both plots, but only derive constants from
the nonlinear fit.
BTW, it would be much easier to read your posts if you could keep
to standard Usenet convention of <75-78 characters per line.
Keith Rickert | "Imprisoned for a crime I didn't commit.
keith at eve.cchem.berkeley.edu | Attempted murder. Now honestly, what is that?
rickertk at netcom.com | Do they give a Nobel Prize for attempted
keith at imppig.caltech.edu | chemistry?" Sideshow Bob, The Simpsons
More information about the Proteins