On Tue, Feb 17, 2009 at 11:08 PM, Sébastien Morin <sebastien.morin.1@xxxxxxxxx> wrote:
Hi, For CPMG, R2eff values will be calculated from: R2eff = ( 1/delayT ) * ln ( Icpmg / Iref ) There is an error on Icpmg and Iref values ; this can be obtained from duplicate measurements.
This could also be done by measuring the RMSD of the baseplane noise. Do you know how errors are then propagated to R2eff? I'll talk to Michael about this too. There could be a direct formula based on Gaussian distributions and Icpmg, Iref, sigma_cpmg, and sigma_ref. Or it could be Monte Carlo simulations (which would be the same as bootstraping as this is an exact calculation and not non-linear curve fitting via optimisation).
For other relaxation dispersion experiments (e.g. R1rho), R2eff values could be extracted from curve fitting of exponential decays. Errors could also be obtained from duplicate measurements.
For this, see help(spectrum.error_analysis). These resultant peak intensity errors are then propagated to the R2eff via Monte Carlo simulations.
For R2eff direct reading, what relax should do is simply store the values with their respective CPMG frequency (for CPMG data) and spectrometer frequency. Then, relax should pass these informations to the code so curve fitting can proceed. What's proposed at: https://mail.gna.org/public/relax-devel/2009-02/msg00009.html (Message-id: <7f080ed10902170945q744a8328x8bc6233131e5d6d3@xxxxxxxxxxxxxx>) seems fine as long as there are no differences between analyses proceeding from intensities (both Icpmg and Iref) or directly from R2eff, i.e. as long as data can be shared easily within relax. I am not sure, however, about how this should be handled... but the way it is done for intensities seems quite versatile and could allow the flexibility needed.
I think I covered this here: https://mail.gna.org/public/relax-devel/2009-03/msg00000.html (Message-id: <7f080ed10903050214x2fd9cfdas6a3a048f724624d5@xxxxxxxxxxxxxx>). Regards, Edward