rank

The PDF modeling process estimates the parameters for a set of parameter estimators. Each estimator is unique in the set of estimators that are processed. Each estimator produces scores on its quality of fit for the distribution. The scores are combined into an overall score for the estimator.

The key thing to remember is that different estimators can be supplied to estimate the parameters from the same distribution family. This is because there can be many different algorithms to estimate the parameters associated with some distribution. For example, the gamma distribution's parameters can be estimated using a method of moments algorithm or a maximum likelihood estimation algorithm. Thus, the scoring results are about the estimators, not about specific distributions unless the set of evaluated parameter estimators do not have multiple algorithms for the same distribution.

Therefore, the ranking of the scoring model is about the estimator used.

A rank of 0, means that the estimation result estimationResult was not found in the results. The returned rank {1, 2, ...} is dependent on the number of estimators that were fit within the PDF modeling process. A rank of 1, means that the estimator associated with the estimation result was the recommended estimator (top ranked) based on the scoring model. Thus, lower ranks imply better estimation fit.

A rank of 0, means that the random variable type rvType was not found in the results. The returned rank {1, 2, ...} is dependent on the number of distributions that were fit within the PDF modeling process. A rank of 1, means that the distribution was the recommended distribution (top ranked) based on the scoring model. Thus, lower ranks imply better distribution fit. The returned rank is the first instance of the random variable type.