## More info on Monte Carlo simulations

The synthetic data sets (of R1, R2 and NOE
values) are created by randomly picking new values
on a Gaussian distribution (with a width given by the experimental
uncertainties) around the values predicted by the optimized value(s) of the
fitted model-free parameter(s) (or possibly around the experimental
values). Model-free parameters are then fit to these synthetic data
sets, and the resulting width of the distribution for each fitted
model-free parameter is a measure of the error in that parameter.

Note that the ModelFree program allows us to use one set of model-free
parameters for generating the predicted relaxation parameters (around
which the synthetic data are distributed), and a
second, different set of model-free parameters (specified by the -s flag)
for fitting to the
synthetic data. This feature is used in evaluating the F-statistic for
comparison of two different models.

#### References

Palmer, A. G., Rance, M., & Wright, P. E. (1991) J. Amer. Chem. Soc. 113,
4371-4380.
Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. (1992)
"Numerical Recipies: the art of scientific computing". Cambridge University
Press, p. 684 ff.

### Hierarchical links

Back to the Model Selection page
Palmer Group Home Page

Updated 9/23/97 by Arthur G. Palmer
(agp6@columbia.edu)