By Group Analysis with Boosted Trees in Workspace - Statistica General Discussion - Statistica - Dell Community

By Group Analysis with Boosted Trees in Workspace

By Group Analysis with Boosted Trees in Workspace

This question is answered

Is there a way to run a By Group analysis with Boosted Trees regression in a workspace?

Verified Answer
  • There is currently no by group option for Boosted Tree regression node in a workspace mode. But we have already logged a design request with ID104198 and it will be considered in the future release of Statistica. 

    In the interactive mode, you can run boosted regression methods by groups with information in KB article "By Group option in the interacive anlaysis".

All Replies
  • There is currently no by group option for Boosted Tree regression node in a workspace mode. But we have already logged a design request with ID104198 and it will be considered in the future release of Statistica. 

    In the interactive mode, you can run boosted regression methods by groups with information in KB article "By Group option in the interacive anlaysis".

  • Here's another related question:  In the Boosted Trees dialog box you get from the Menu bar (as opposed to the workspace node), it possible to select the minimum cases for a stopping parameter on a percent basis (like the workspace node) as opposed to an absolute number of cases?

  • In interactive module, the minimum cases can only be specified as n rather than a percent basis.Do you want to have the capability of specifying as percent? Could you tell more about why would you think this would be nice for your use case?

  • The percent basis would be nice because at the end of a boosted tree analysis initiated from the menu bar, the option is given to run a By Group analysis.  However, when the original data set is split into groups, the absolute number of cases may no longer make sense, whereas using a percentage would scale the minimum cases for the new smaller groups.

  • That's a good use case. I log a design issue for this with ID 104763 and our production will evaluate to consider including it in the future release of Statistica.