CONSTRUCTING A SAMPLE SELECTION MODEL OF HIGH RISK YOUTH

申请人摘要： 这项研究的重点是测试技术，以促进 确定参与毒品活动的高危青少年，学校 脱离和自杀危险行为。该提案针对的是 准确识别高危青年的基本和务实需要 在不产生额外数据收集成本的情况下进行预防 或者是筛查年轻人的复杂程序。开发和测试 准确高效地找到高危青年的简单模型 在学生中是一个被忽视的研究领域；然而，这是 这是任何预防性干预研究的关键第一步。 研究的目的是(1)测试现有的高危青年选拔 目前用于识别高危青少年学业失败的模型， 并因此，毒品参与；(2)评估替代模型 最大化的选择 数据收集效率和分类精度；(3)探索 模型筛选风险程度的能力；以及(4)测试 针对特定种群亚组的选择模型：a)男性/女性， B)少数民族和c)以年龄为基础的小组(14-15岁和16-18岁)。 这项研究将使用现有的数据库；即以前 收集的数据来自(五)年的学区记录和 学生个人调查，包含有关毒品的基本资料 参与、在校表现和自杀危险行为。措施 这些行为具有很高的信度和效度，并构成了 对学生人口进行分类。研究设计要求 将数据拆分为子样本，使用部分数据派生 选择模型(主要基于分类和回归树， 购物车、程序)，然后使用剩余的样本来测试 派生模型的准确性。这种设计类似于典型的 对半设计。购物车程序专门设计用于 (1)从较大的变量集合中找到有限数量的分类器 以及(2)最小化分类中的错误率。 在学区文件中系统收集的变量 将被用作个体的独立因素或量词 病例；这些分类器是众所周知的风险因素的代表。 所表达的兴趣是在现有数据集中工作，通常 在学校环境中有系统地提供，以产生一种 精准选拔高危青年。 这项研究的价值在于测试和确定一种简单的方法来 对典型场所的风险状况进行分类以进行预防干预 项目，学校。如果这样的程序成功，高风险 青年样本选择模式将为未来提供基石 研究，提供了一种简单而实用的手段：(1)评估 (2)将高危青年与高危青年联系起来 (3)在全校范围内实施干预计划 设置。

APPLICANT'S ABSTRACT: The focus of this study is on testing techniques that facilitate the identification of youth at-high risk for drug involvement, school disengagement, and suicide risk behaviors. The proposal addresses a fundamental and pragmatic need to identify accurately high-risk youth for prevention efforts without creating costs of additional data gathering or complicated procedures for screening youth. Developing and testing simple models for accurately and efficiently finding high-risk youth among school populations is a neglected area of study; and yet, this is a critical first step in any preventive intervention research. The research aims are to (1)test an existing high-risk youth selection model currently used to identify youth at-high risk of school failure, and thereby, drug involvement; (2)evaluate alternative models of selection which maximize efficiency of data collection and accuracy of classification; (3)explore the ability of the models to screen for degree of risk; and (4)to test models of selection for specific population subgroups: a)male/female, b)ethnic minorities and c)age-based subgroups (aged 14-15 & 16-18 years). An existing data base will be used for this study; i.e., previously collected data from (five) years of school district records and individual student surveys which contain basic information on drug involvement, school performance, and suicide risk behaviors. Measures of these behaviors have high reliability and validity and form the basis for classifying the student population. The research design calls for splitting the data into subsamples, using part of the data to derive the selection models (primarily based on Classification and Regression Tree, CART, procedures) and, then, using the remaining samples to test the accuracy of the derived models. This design is analagous to typical split-half designs. The CART procedures are specifically designed to (1)find a limited number of classifiers from a larger set of variables and (2)minimize error rates in classification. Variables that were systematically gathered in the school district files will be used as the independent factors or classifiers of the individual cases; these classifiers stand as proxies for well-known risk factors. The expressed interest is to work within an extant dataset typically and systematically available in a school setting to produce a model that accurately selects youth of high-risk status. The value of this study is to test and to identify a simple means for classifying risk status in a typical site for preventive intervention programs, the schools. If such a procedure is successful, the high-risk youth sample selection model will provide a building block in future studies, providing a simple and pragmatic means for : (1) assessing prevalence of risk in school populations; (2) linking high risk youth to interventions and (3) implementing intervention programs across school settings.