Tree module

[David Kantor]

Hello All,

Thanks to Kit Baum, the tree module is available in SSC.

This facilitates the modeling of decision trees, and the evaluation of payoff variables toward the roots of trees according to given probability values.

Facilitates the process of...

• specifying the structure of one or more decision trees, and establishing a set of payoff variables;
• populating the initial (terminal node) payoff and probability values;
• evaluating the trees -- propagating payoff values toward the roots, according to the probability values.

The advantage is that you can develop the initial payoff and probability values in Stata, and transfer them directly into the trees. Prior to developing this facility, my colleagues had been building the trees in Excel, and copying values from Stata into the Excel tables, which is labor-intensive and does not lend itself to easily evaluating variations and alternative analyses. (There are other options besides Excel, but they still require copying values in from whatever process generates them.) The tree program enables you to quickly evaluate variations and alternative scenarios.

This has been used in economic evaluation of healthcare procedures.

Additionally, it can be used to model hierarchical structures, calculating sums and means toward the roots.

Tree structures are recorded in the current dataset; each node is an observation. If you develop initial payoff values or probabilities in Stata, then you need to swap between the value-development data and the tree data. That's not difficult, just a consideration to be aware of. You can probably expedite the process by using frames, but I have not tried that.

Requires Stata 14.2.

ssc desc tree
ssc install tree

I'll be curious to know whether anyone finds this interesting or useful.

Thanks / cheers,
--David