Analytics

When implementing a PDT run, various analytics are generated at the end of each generation. Below we give a brief description of each of the graphs available on the Analytics tab (visible only after a run has been started).

Response Barplot Time Series

The responses for each generation are sorted, and presented one after the other. Standard error bars are given at the top of each colored bar.

Response Barplot by Generation

Detail of the previous graph, giving a separate graph of sorted responses for each generation.

Overlapped Response Barplots

A summary of the data in the previous two graphs, sorted response barplots for each generation overlapping each other, in different colors.

Response Boxplot Time Series

A boxplot representation of the distributions of response values for each generation, presented one after the other.

Response Histogram (All)

The response of all experiments in all generations.

Response Histogram (by Generation)

Response histograms, generation by generation.

Evolution Time Series

For each experimental parameter (i.e., each of the variables specified in the experimental space definition), the relative representation of that parameter is represented for each generation. Relative representation is measured by a percentage, namely the percentage of experiments for a given generation having a particular parameter at a particular value.

In the course of evolution, parameter values that have low representation tend to be those that do not contribute to high response; those with high representation tend to contribute more.

Conditional Inference Tree

This plot uses the R package CTree. CTree is a non-parametric class of regression trees embedding tree-structured regression models into a well defined theory of conditional inference procedures.

For data sets with relatively simple structure, this tree will have only a few nodes. (NB: this may be the case with the synthetic data in the demo runs.)