Back to TotalFit Methods

Evaluation of parameter correlations

Correlation plots show you directly how parameters compensate for each others fluctuations and also display correlation coefficients. The value of R_cutoff gives a minimum correlation coefficient between parameters to have the plot created by plot_all_correlations() method. However, correlation coefficient sometimes takes crazy values therefore I usually set R_cutoff=0. A cell array 'correlations' contains a list of parameter names whose correlations that are larger than R_cutoff.

IMPORTANT NOTE: MATLAB has very limited memory for figures and would freeze if number of displayed figures exceeds some 40. Global fitting runs or fitting with very large models yield large numbers of parameters. In such situations one should turn off automatic display of all correlations and histograms and switch to display of selected graphs only: set show_all_correlation_plots to 0 and use a method plot_correlation() for a specific parameter pair and plot_specified_correlations() for a set of parameter pairs.

 

For a specific usage see TotalFit.m