Relaxation dispersion model elimination

Relaxation dispersion models will often fail. This may be due to data quality and quantity issues, inherent instability in certain models, or the use of analytic models outside of the range of their defined viability. Model elimination is therefore required to remove these failed models prior to model selection, as failed models will often fit the experimental data statistically better than non-failed models. The user function eliminate (see Section 17.2.41 on page [*]) is used to remove the failed models. Model elimination was implement in relax as described in:

The following hard coded rules are used to eliminate models:

p_{\textrm{A}}\leqslant 0.501, \\
...slant 0.999, \\
\tau_{\textrm{ex}}\geqslant 1.0.

If a parameter falls outside of these limits, the entire spin cluster will be deselected. When not using the auto-analysis, custom model elimination rules can be defined and used with the eliminate user function.

The relax user manual (PDF), created 2020-08-26.