Technical Issues Knowledge Item 07: Results from HRSA/HAB's SPNS Innovative Models of Care

Knowledge Item: CA-Technical Issues-07
Missing Data Categories in Data Partitioning Tree (or CHAID) Analyses

CHAID is a data modeling method that has been used extensively in this Knowledge Base. This Knowledge Item (Extended Results) looks at the consequences, for a number of data sets, of letting "missing value" categories merge with the observed value categories. Three methods are contrasted: a) letting the CHAID algorithm generate variable splits and merges; b) not allowing CHAID to merge categories on any possible predictor variables; c) using a hybrid method wherein the optimal split and merge categories are used to determine the importance of the variable as a splitter, but then adjusting the split so that the missing value category is kept separate.

In most cases, the merging of the missing values with the observed values is inconsequential and tends to strengthen the models. The interpretation of the split, however, needs to be adjusted slightly.

The hybrid splitting procedure is developed and illustrated here.

The Extended Results section shows two key examples and a description of the modified splitting algorithm. The Additional Statistics section shows additional contrasts of the three methods.

Automatic Algorithm (limited to 2 splits)

Modified Algorithm (limited to 2 splits)

Click graphic to expand. (IE 6 users may also have to click the graphic expansion button in the new window.)

More Information: CHAID and CHAID Diagram

Knowledge Item Citation: Huba, G. J., Panter, A. T., Melchior, L. A., and the HRSA/HAB SPNS Cooperative Agreement Steering Committee (1998-2001). Knowledge Item: CA-Technical Issues-07 from HRSA/HAB's SPNS Cooperative Agreements on Innovative Models of Care, The Measurement Group Knowledge Base on HIV/AIDS Care, Online at www.TheMeasurementGroup.com.

Last Updated: March 25, 2005; data through June 15, 1999; analyses conducted April - June 2000.

Knowledge Base Citation: The Knowledge Base and this Knowledge Item were designed and authored by G. J. Huba, Ph.D.; in collaboration with Lisa A. Melchior, Ph.D.; A. T. Panter, Ph.D.; and the staff of The Measurement Group. Cite this work as "Huba, G. J., Melchior, L. A., and Panter, A. T. (1998 - 2001). The Measurement Group Knowledge Base on HIV/AIDS Care. On the World Wide Web: http://www.TheMeasurementGroup.com."

Questions or Comments: Contact The Measurement Group.

Use of Knowledge Base Information: Acceptable Uses and Limitations.

Collaborators from Participating Projects: Cooperative Agreement Steering Committee 1999.

Participating Projects: This Knowledge Base is based on the service delivery experiences of 27 Cooperative Agreement Projects on Innovative Models of HIV/AIDS Care. These projects and the Evaluation and Dissemination Center which produced this Knowledge Base were funded by the Health Resources and Services Administration (HRSA), HIV/AIDS Bureau (HAB) as Special Projects of National Significance (SPNS) between 1994 and 1999. Click the Model Programs button above for descriptions of the projects that contributed to this specific Knowledge Item, a list of key staff, and project grant numbers.

Why This Evaluation was Conducted: Editorial.

More Information: Design of this Knowledge Base.

Recommended Citation Format for Web Materials: American Psychological Association Publication Manual Section, Revised 2001.

Work on the Knowledge Base and the cross-cutting evaluation was supported in part by Grant Number 5 U90 HA 00030-05 from the Health Resources and Services Administration (HRSA), HIV/AIDS Bureau's (HAB) Special Projects of National Significance (SPNS). The contents of this Knowledge Base are solely the responsibility of The Measurement Group and do not necessarily represent the official views of HRSA or HRSA/HAB's Special Projects of National Significance nor may they represent the positions of the individual grantees whose projects are included in the cross-cutting evaluation.