Fitting results

Three dimensional plot of raw data.

The command window gives useful information about the progress of the fit, and of the final fit results. It also indicates if convergence problems have occurred.

It is also possible to visualise the fitting progress by plotting the sum of squares for each iteration. The fit can be interrupted by pressing the shown pause and stop buttons, or alternatively stopped by pressing control+C when the command window is selected.

Fit plotted on top of raw data (top panel). Patches on top of the grid (the raw data) are an indication that that region of the raw data is not perfectly fitted (the fit is symbolised by the continous color plot, a full surface without grid). The bottom panel shows the concentration profiles for each time constant fitted to the data. The units of the time scale are the same as those defined in your input data. The sum of squares is also given in the title of the top panel, and can be used as quality of fit indicator.

Time traces and their fit. The raw data is drawn in black, the fit in blue. Up to 6x40 traces can be plotted (6 figures that contain 40 small plots each). The plot title is the index (the column index of your data file, representing a channel for instance), or your calibrated value. In this case the fit traces the data nicely. Note also that the time base is logarithmic. Therefore, if the data contains negative time points, these are not shown. Make sure time zero in your dataset is before the appearance of any signal.

It is also possible to plot all traces on a linear time base (see the plotting options panel). However, the traces with the new time base are only regenerated after having completed a new fit (not by pressing Replot fit).

Time-independent spectra, corresponding to the estimated time constants. The used model is shown (Parallel), the time constants (in the legend). It is also possible to estimate the confidence intervals for the spectra (shown in the plot).

The fit quality can be explored by clicking once inside the surface plot to activate a cross-hair, and a second time to select a point inside the surface plot. The selected point (the center of the cross-hair) determines which time trace and which spectrum to plot individually. The fit (blue line) is plotted together with the raw data (black line) for comparison.

Residuals in time and over all pixels for investigation of the fit quality. The sum of squares is also given. Structure (i.e. non-random features) in the time residuals (top panel) such as oscillations indicate that the fit could be improved. The spectral residues (bottom panel) shows then the spectral signature of the misfit in time. In this case the dip at 1.3 seconds (the maximum in the top panel) is the dominant contributor to the spectral signature in the bottom panel. The size of the spectral residue is very small compared to the actual signal, so it seems that this fit is rather adequate, and the residuals are mainly caused by the accuracy of the solver.