News archive for September 2008

DataModeler Release 7.0

Thursday, September 18, 2008

Release 7.0 has several minor changes:

  • Added tooltips to EnsemblePredicitonPlot and similar functions are useful for exploring outliers;
  • Adapted VariablePresenceMap to accept lots of variables;
  • Extended SplitInputResponseData (useful when the targeted response is buried in the middle of a data matrix);
  • Modified EnsemblePredictionPlot, EnsembleResidualPlot, ModelPredictionPlot and ModelResidualPlot so that tooltips may be displayed for each point. Setting ToolTipFunction -> None does not display tooltips, a setting of Automatic (the default behavior) will display the record number and ToolTipFunction -> (pureFunc[#]&) will receive the index number as its argument. (Some obsolete option settings were also deleted.)
  • Modified VariablePresenceMap so that setting ModelingVariables -> Automatic will suppress any variables which are not present in the selected models;
  • Modified ParetoFrontPlot, ParetoFrontLogPlot, ResponseSurfacePlot and ModelSelectionReport so that options are automatically supplied to the specified ToolTipFunction. The implication here is that DataVariableLabels may be supplied and the resulting ModelExpressions in the ParetoFrontPlot and ParetoFrontLogPlot will make the appropriate substitutions;
  • Extended the SplitInputResponseData function so that one or more columns may be identified as a response at an arbitrary column(s) within the supplied matrix. The extracted inputMatrix may also be a (potentially reordered) subset of the initial matrix. The default behavior is the same as the previous implementation; however, now the extracted response no longer needs to be the last column in the data array;
  • Modified MergeInputResponseData so that a response matrix can be supplied as well as vector. In both cases, the input is augmented with additional columns at the end;
  • Fixed a bug in ModelExpression where supplying a list of strings for the DataVariables option would cause problems;
  • Fixed a bug in ImportDataMatrix. Now supplying 0 or 0.0 for DeleteEmptyColumns, DeleteNonNumericColumns or DeleteNonNumericRows is equivalent to a setting of True (i.e., zero tolerance for incomplete data). Also relaxed the structure requirements for ImportDataMatrix[header, data, opts]; now the header simply needs to be a list and the data must be a 2D structure (not necessarily a matrix and not necessarily only having a Depth of three).