Learning Objective #1: determine methods to achieve statistical control of operationalized variables in intervention studies of subjects from vulverable popualtions.

Learning Objective #2: validate study analysis and study conclusions through the effective use of statistical analyses.

---

Purpose: The purpose is to examine control of intervention study variables in studies that include subjects from vulnerable populations such as HIV.

Objectives: 1.To examine study samples for normal distribution, univariate and multivariate outliers; 2.To identify outliers and needed data transformations 3. To achieve control of variables for valid analyses; 4. To select statistical tests to achieve control of moderator and extraneous variables.

Methods: Study variable were examined for descriptive statistics on dependent variables. Analyses of data using the proposed study design did not incorporate subjects' HIV status resulting in no significant differences between study groups. Analyses were then conducted accounting for subjects' HIV status because of correlations between dependent variables with HIV status.

Analyses: Careful examination of study variables when subjects are members of vulnerable populations is critical to the validity of data analyses. Failure to carefully examine data to determine aberrations or failure to meet the statistical assumptions for the use of inferential interval level statistics could result in Type 1 or Type 2 errors. In studies where moderator variables such as HIV status are not controlled the analyses could lead to erroneous conclusions Careful analysis of data may lead to the identification of variables within groups that may exert pull on variables resulting in the clustering around the mean ultimately masking between group differences on variables.

Analyses incorporating HIV status resulted in statistically significant differences supporting the effectiveness of the intervention for HIV negative subjects.

Evaluation: Appropriate use of statistics is essential to the validity of study results. Scrutiny of data with statistical consultation is important to the research process. Analyses using different statistical techniques can result in very different outcomes on the same data.

Implications: Investigators must carefully examine data prior to proceeding with inferential analyses. Consultation is an important consideration even for experienced investigators.