*** For details on how to be removed from this list visit the *** *** CCP4 home page http://www.ccp4.ac.uk ***
All - The statistical standard uncertainty SU (previously known as "estimated standard deviation" or "standard error") is NOT the same thing as the "RMS deviation from the mean", and cannot be computed from the Fourier components. The RMSD (and the Fourier components) include contributions from all grid points in the map, including the ones in the atomic densities, which can hardly be described as "error". To get the true uncertainty you have to exclude all these points, which can obviously only be done in real space. Hence the SU will always be smaller than the RMSD, and there will indeed be significant features in the map < 1 sigma, if sigma is incorrectly equated to the RMSD instead of the SU. The extends program makes an attempt at estimating the true SU, though probably not a very rigorous one. The correct procedure would be to estimate the SU from a normal probability plot. This involves ordering the deviations by their (signed) magnitudes, then comparing them with the theoretical values assuming a normal distribution of errors. The outliers due to the signal will then appear at the extremes of the plot, which can excluded from consideration, and the true SU due to the random errors can be estimated from the gradient of the central part. This is something I've been meaning to do in extends for a while (but then how many people other than me are using extends anyway?). The difference between SU and RMSD is most obvious in a 2mFo-DFc map (or indeed any type of Fobs map), because a large fraction of the map is then signal and it's impossible to estimate a sensible SU. As others have pointed out before, the RMSD of such a map depends largely on the solvent content of the crystal, not the true error! This is not the first time RMSD and SU have been confused, there was a recent discussion concerning the RMSD of bond lengths from the dictionary values after refinement. In this case the RMSD will always be less than the rms SU of the bond lengths (which is known from the small molecule data) because of the reduction in the degrees of freedom by the number of refined parameters (indeed there's nothing to stop the RMSD being zero if the NDF is zero!). -- Ian > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On > Behalf Of Bart Hazes > Sent: 12 August 2005 00:27 > To: Artem Lyubimov > Cc: [email protected] > Subject: Re: [ccp4bb]: Quantifying density peaks in MAPMAN > > *** For details on how to be removed from this list visit the *** > *** CCP4 home page http://www.ccp4.ac.uk *** > > > Artem Lyubimov wrote: > > *** For details on how to be removed from this list visit the *** > > *** CCP4 home page http://www.ccp4.ac.uk *** > > > > > > Hi all, > > > > Pardon the non-CCP4 question. I'm using MAPMAN to > calculate electron > > density peak heights at atomic positions, but I cannot > figure out how to > > determine the standard error for this calculation. How do > I do this? > > > > Thanks! > > Art > > MAPMAN is written by Gerard so there are no "standard errors" > only the > very highly specialized errors that nobody but Gerard has mastered ;) > > I expect what you refer to is the standard deviation (or RMS > value) of > the map so you can state that a peak is x sigma above > background. When > you created the map the RMS was probably listed in the log > file. MAPMAN > may also have its own command to determine this value. Interestingly, > you can calculate the RMS value right from the Fourier coefficients > without having to calculate the map. SFTOOLS does this (MAP > RMS 1 gives > the RMS for the map based on amplitudes stored in column 1). > > In case MAPMAN calculates the density value by interpolation then a > small estimation error is added. For the puritan (or > paranoid) you can > use direct Fourier summation to get the exact value. If you belong to > either category then I can give you a copy of my HYDENS > program but that > uses Fourier summation. However, the error in your data and phases is > almost certainly a lot larger than the interpolation error, > especially > if you calculate the map on a relatively fine grid. > > I hope this answers your question. If you actually meant > something else > then please clarify. > > Bart > > -- > > ============================================================== > ================ > > Bart Hazes (Assistant Professor) > Dept. of Medical Microbiology & Immunology > University of Alberta > 1-15 Medical Sciences Building > Edmonton, Alberta > Canada, T6G 2H7 > phone: 1-780-492-0042 > fax: 1-780-492-7521 > > ============================================================== > ================ > > > ********************************************************************** CONFIDENTIALITY NOTICE This email contains confidential information and may be otherwise protected by law. Its content should not be disclosed and it should not be given or copied to anyone other than the person(s) named or referenced above. If you have received this email in error, please contact [EMAIL PROTECTED] **********************************************************************
