PFChrom v5 Documentation Contents AIST Software Home AIST Software Support
When you click the Peak Fit button in any of the fitting options, you will see the following dialog:
Peak Fit Strategy
The strategies are as follows:
Fit with Full Data
With this option a single fit is made as was done in all prior versions of the product.
Fit with Reduced Data Prefit
To speed up the fitting, this option employs an initial fit with a lower density of data. This usually results in the estimates being much closer to the global optimum before a full data fit is initiated.
Fit with Reduced Data Prefit, 2 Pass, Lock Shared Parameters on Pass 1
In this fitting strategy, the reduced data prefit is done in two steps. In the first step, all parameters marked as shared (typically the ZDD and IRF parameters) are locked and not allowed to vary. This allows the main model parameters to be more accurately determined before the higher order ZDD or the various IRF parameters are allowed to vary. If you are fitting a ZDD or IRF model, we recommend that you employ at least this sophistication of strategy. IRF parameters can readily assume values which can invalidate fits when the primary model parameters, especially the a3 distortion or asymmetry, is not yet close to an optimum value. One does not want the IRF or ZDD parameters competing for this intrinsic asymmetry.
Fit with Full Data, Cycle Peaks
With this option, the full data is used, but each peak will see a separate fitting step where it is the primary focus. This allows small amplitude peaks to be successfully and accurately fitted alongside much higher amplitude peaks whose fit errors can often 'wash-out' much smaller peaks.
Fit with Reduced Data Prefit, Cycle Peaks
In this option, a reduced data density is first used in the cycle peaks strategy. This is then followed by a full data fit where this cycle peaks strategy is repeated. This is usually beneficial in fitting time if there is sufficient information in the reduced data to accurately model the individual peaks.
Fit with Reduced Data Prefit, Cycle Peaks, 2 Pass, Lock Shared Parameters on Pass 1
This combines all strategies to achieve the best fitting speed while managing higher moment and IRF fits, and accommodating very low area components. We use this strategy almost exclusively when fitting multiple peaks with a ZDD and/or IRF-bearing model.
The strategy options are as follows:
Fit using Sequential Constraints (to nudge fits to global minimum when initial estimates weak)
This adds the intelligence of restarting the above strategies at a specified iteration count, resetting the values of the general constraints in the Peak Fitting Preferences. This is what you would see in prior versions when manually restarting a fit after a first convergence was attained. This allows a fit to be more gracefully and effectively nudged to the global minimum.
Fit using Separate Segments (for mix if baseline resolved peaks and segments of overlapping peaks)
The Baseline Segments checkbox in the placement screen only places the peaks as if clusters exist in baseline-separable segments. To actually perform a segment-based fit, to fit these peaks as if they were in entirely distinct baseline-separated sections, you will need to check this option.
Offer Simpler Models when Appropriate (assures all parameters statistically significant)
This option is experimental and is available only when fitting a single data set. For certain models, this will downgrade the complexity of a model and repeat the fitting until all parameters in the overall model are statistically significant (not overspecified).
The segmented fit addresses the reality of incomplete data definition on overlapping peaks. You should thoroughly explore this particular innovation in PFChrom if you routinely encounter peaks which are not baseline resolved.
Although a fit takes longer, we recommend the Fit with Reduced Data Prefit, Cycle Peaks, 2 Pass, Lock Shared Parameters on Pass 1 strategy and the Fit using Sequential Constraints (to nudge fits to global minimum when initial estimates weak) option, especially when fitting a ZDD (zero-distortion density) higher moments fit, or if an IRF (instrument response function) is fitted.
In such a fit, there may be hundreds of separate independent fits, some on subsections of the data, some on reduced density data, some with specific parameters locked, and in a sequential accommodation of the constraints. These settings are our 'go-to' choice for automated fitting.
Real-Time Graphical Fitting
Click OK to initiate the fitting. This will open the real-time graphical fitting window.