Note on the scaling of predictor variables: Unless otherwise noted, the indicators coded as "current-prior-no" were typically treated as nominal variables in the exploratory CHAID analyses conducted here. That is, rather than saying that a variable [such as coding use of heroin] has three categories and the relationships to the dependent variables should be no use > prior use > current use [or conversely no use < prior use < current use] we allowed the three categories to be ordered in the data as empirically determined to be appropriate. In many cases we have done alternate analyses considering these categorical variables [blue above] to be ordinal and the results are essentially the same. The choice of the nominal scaling was favored in that it was not always clear that individuals with prior risks would be treated by programs in a way intermediate to those with no risk and current risk. At any rate, in most cases there are no substantive differences attributable to the variable measurement issue. Note also that in all of our CHAID analyses on these variables, missing values on the service need and vulnerability indicators were considered to be a potentially valid data elements in that if the indicators were not known by the service providers, then service histories might be different.

Note on CHAID models: CHAID is a useful method of summarizing data, and can show major natural divisions of the sample by various defining variables. It must be recognized, however, that CHAID is analogous to a "forward" stepwise regression analysis and has all of the possible inferential difficulties of such stepwise regression methods. The statistical significance tests are sequential ones dependent on prior splits of the sample. In many cases, the models presented should be considered as suggestive, but not absolutely definitive as there may be alternate models that also fit these data in a statistically or theoretically acceptable manner. This model may have been manually altered very slightly from the automated CHAID modeling trees; specifically, categories for "missing values" may have been separated (or re-split) from categories for actual values with which they were statistically merged if the authors judged this would give a more clear interpretation of the data; the separation may result in a "missing category" with only a few cases that could be statistically merged with one of the other categories. [In those cases where the "missing value" category is combined with actual values, it was judged that the automated split was a better representation of these data.] The use of Bonferroni confidence intervals to correct for the potentially large number of statistical tests in this model building method and the use of more stringent alpha levels results in relatively conservative data representations. All patient-client model analyses were conducted by the senior author [GH] so that consistent data fitting techniques and judgments would be employed in the different areas studied. In many cases, alternate models are presented so that the viewer can judge the appropriateness of one or more ways of looking at the same data.

Appendix: More Detailed Description of How Selected Variables are Coded

Notes illustrating specific issues.

Note 1. Coding Racial-Ethnic Group from original codes.

First Generation Analyses [June 15, 1999]: In most cases these analyses were or will be replaced with second generation predictors.

In the first generation CHAID analyses reported here, in addition to gender, indicators of service needs and vulnerabilities were tested as potential predictors of the dependent variable.

The potential predictors were

• under age 21 [coded as yes, no]
• person of color [coded as yes, no]
• primary language other than English [coded as yes, no]
• children requiring care [coded as yes, no]
• education of less than 12 years [coded as yes, no]
• not working [coded as yes, no]
• dependent on public support for medical services [coded as yes, no]
• current or prior alcohol problem [coded as current, prior, no]
• used or had used heroin [coded as current, prior, no]
• used or had used crack [coded as current, prior, no]
• used or had used another illicit drug [coded as current, prior, no]
• were or had been involved with the criminal justice system [coded as current, prior, no]
• engaged in current or prior sex work [coded as current, prior, no]
• without stable housing [coded as yes, no]

As the analyses for this Knowledge Base were conducted over an 18-month period, two major variations of the labeling of the service needs and vulnerabilities indicators were used (with certain other minor variations occasionally used). The variables and data are the same in all analyses; the labels were changed to make some of the analyses (especially those done near the end of the analysis cycle) more self explanatory in the case of complex CHAID diagrams and other analyses. In some cases, the CHAID diagrams in a Knowledge Item will label the variables using Variation 2 while a series of pie diagrams may label them using Variation 1. [Note: the variable labeling-coding does not affect any of the statistical results, just the variable labeling within tables or diagrams.]

Note on the scaling of predictor variables: Unless otherwise noted, the three category variables were typically treated as nominal variables in the exploratory CHAID analyses conducted here. That is, rather than saying that a variable [such as using heroin coded] has three categories and the relationships to the dependent variables should be no use > prior use > current use [or conversely no use < prior use < current use] we allowed the three categories to be ordered in the data as empirically determined to be appropriate. In many cases we have done alternate analyses considering the three category variables [blue above] to be ordinal and the results are essentially the same. The choice of the nominal scaling was favored in that it was not always clear that prior users of drugs would be treated by programs in a way intermediate to nonusers and active users. At any rate, in most cases there are no substantive differences attributable to the variable measurement issue. Note also that in all of our CHAID analyses on these variables, missing values on the service need and vulnerability indicators were considered to be a potentially valid data elements in that if the indicators were not known by the service providers, then service histories might be different.

Note on CHAID models: CHAID is a useful method of summarizing data, and can show major natural divisions of the sample by various defining variables. It must be recognized, however, that CHAID is analogous to a "forward" stepwise regression analysis and has all of the possible inferential difficulties of such stepwise regression methods. The statistical significance tests are sequential ones dependent on prior splits of the sample. In many cases, the models presented should be considered as suggestive, but not absolutely definitive as there may be alternate models that also fit these data in a statistically or theoretically acceptable manner. This model may have been manually altered very slightly from the automated CHAID modeling trees; specifically, categories for "missing values" may have been separated (or re-split) from categories for actual values with which they were statistically merged if the authors judged this would give a more clear interpretation of the data; the separation may result in a "missing category" with only a few cases that could be statistically merged with one of the other categories. [In those cases where the "missing value" category is combined with actual values, it was judged that the automated split was a better representation of these data.] The use of Bonferroni confidence intervals to correct for the potentially large number of statistical tests in this model building method and the use of more stringent alpha levels results in relatively conservative data representations. All patient-client model analyses were conducted by the senior author [GH] so that consistent data fitting techniques and judgments would be employed in the different areas studied. In many cases, alternate models are presented so that the viewer can judge the appropriateness of one or more ways of looking at the same data.

Appendix: More Detailed Description of How Selected Variables are Coded

Last Updated: January 02, 2008; data through June 15, 1999; analyses conducted November 1999 - May 2000 .