On 17 January 2012 12:43, Paul Emsley <[email protected]> wrote:
> On 17/01/12 08:33, Dayana Nisbar wrote:
>>
>> How to change the map level from rmsd to sigma?
>
>
> You can't. Electron density levels should not be expressed in terms of
> sigma. Sigma refers to probability distributions - and an electron density
> map is not one of those. I used to not understand this and there may be
> some references to "sigma" in the code/interface which I need to clean up.
But an electron density, being an experimentally determined quantity,
has an error and an error certainly does have a probability
distribution. The standard deviation of that error distribution,
which is not the same as and should not be confused with the RMSD, is
what I assume is here being referred to as 'sigma'. The RMSD would be
the s.d. if the electron density were to be treated as a probability
density function (which I agree it isn't); however this pseudo-PDF is
not in any case relevant to the discussion because what we are
concerned about here is the error distribution (characterised by
sigma), not the electron density distribution (characterised by the
RMSD).
When we are judging significance of the (difference) electron density
the relevant statistic is the signal/noise ratio (delta-)rho/sigma,
analogous to I/sigma for the diffraction data (no-one would think of
expressing I in terms of its RMSD!). The s.d. (sigma) of the
difference density can be estimated from a Q-Q ('normal probability')
plot. (Delta-)rho/RMSD has no meaning in terms of significance (it
has no meaning period), so electron densities should not expressed in
terms of RMSD, unless you can legitimately claim that the RMSD is a
good estimate of the s.d., which means that you are really expressing
it in terms of sigma. Also it begs the question "how accurate an
estimate of the s.d. is the RMSD?" and "is a more accurate estimate
available?".
Cheers
-- Ian
Cheers
-- Ian
