SI-CHAID option in Latent GOLD and LG Choice

Extension to multiple dependent variables

Often segmentation is desired that is predictive of not one but multiple criteria. For example, in database marketing, dependent variables might include 1) response to the most recent mailing (responder vs. nonresponder), 2) response to past mailings, 3) the amount spent, 4) profitability, and possibly others. It is now possible to obtain CHAID segmentation trees that are predictive of multiple dependent variable criteria. In addition, these dependent variables may be continuous, ordinal, nominal, or count variables, or any combination of these!

The groundbreaking article “An Extension of the CHAID Tree-based Segmentation Algorithm to Multiple Dependent Variables”, Magidson and Vermunt (2005) shows this is possible. The key is to use latent classes as a proxy for the multiple dependent variables. This can be done with Latent GOLD 4.5 when the dependent variables are used as indicators in a latent class cluster or factor model, or it can be done with Latent GOLD Choice 4.5 when the dependent variables are choices obtained from a discrete choice study.

Each of our flagship modeling tools Latent GOLD 4.5 and Latent GOLD Choice 4.5 provide a direct link to SI-CHAID 4.0. With this option, a CHAID Definition (.chd) file is automatically generated immediately following model estimation which can then be used as input to SI-CHAID 4.0.

To see how this works:

• download the article
• view a tutorial where a CHAID tree is developed that is predictive of 11 dependent variables in conjunction with Latent GOLD 4.5 (SI-CHAID tutorial #4)
• view a tutorial from Latent GOLD 4.5 where a CHAID analysis is used with a discrete factor model to better understand the factors (LG tutorial #4)
• view a tutorial from LG Choice: Tutorial 1A: Using CHAID to Profile Latent Class Segments

---

Latent GOLD and LG Choice are separate, stand alone programs. Learn More