I am using cadetmatch to fit for ka/kd (auto-keq), qmax, and surface diffusivity. The fitting works well and now I am trying to obtain the confidence intervals for these parameters. In the folder of ‘progress’ there are several plots, including corner, corner-min, corner-min-transform, corner-prod, corner-prod-transform, corner-transform, do you know which one I should use to obtain the confidence intervals? And what is the percent of the intervals?
Also, is there a code line I can use to directly obtain the confidence interval values, instead of reading the numbers from the corner plots?
Thanks for your help!
While those corner plots can tell you if a parameter is identifiable they can’t give anything like confidence intervals. In order to get confidence on the parameters MCMC is needed and CADETMatch can automatically do that but it does take a lot of simulations. The documentation is available for it here Error Modeling — CADET-Match
From the command line you can also have CADETMatch generate examples CADET-Match Examples — CADET-Match and there are some MCMC examples in there also.
I don’t know of a shortcut way to do this short of running MCMC because the effects of the input parameters are highly non-linear. If you do run MCMC it should be largely automatic but you will need to set the values in the error model based on your system such as the UV detector noise, how accurately your solutions tend to be mixed, and how accurate the pump flow rate is.
Thanks Bill for your detailed answer! This really helps a lot.