# Sampling Fit

A Sampling Fit is a combination of space-filling DOE method and mathematical model trained by the data generated.

To build an accurate Fit model, it is crucial to have enough DOE runs. Since it is challenging to define the right number of runs in a single DOE, the solution has been to sample with an initial guess to build a Fit model. Then, if more data is needed, perform more sampling until the expected accuracy is achieved.

Sampling Fit remedies this by continuously sampling the design space and building a Fit
at specified, fixed intervals until the required cross-validation R^{2} is
achieved.

Modified Extensible Lattice Sequence (MELS) is used as DOE Method.

Parameter | Default | Range | Description |
---|---|---|---|

Stopping R^{2} |
0.9 | > 0.0 | Cross-validation R^{2} value. Sampling stops as soon as
this value is satisfied. |

Maximum Evaluation Count | 50 | > 0 integer | Maximum number of evaluations allowed. |

Evaluations Per Iteration | 2 | > 0 integer | Controls the fequency of building a fit model to check
cross-validation R^{2} value. |

Sequence Offset | 1 |
Integer 0 to ∞ |
Controls the starting offset for the Modified Extensible Lattice Sequence. For example, a value of 101 starts the generated evaluation points from the 101st point of the Modified Extensible Lattice Sequence. |

Filter Rungs with Bad Runs | On | Off or On | Filters runs with missing/invalid results per response. |