Browsing University of Alaska Anchorage by Subject "research methods"
Now showing items 1-4 of 4
ADAM-Anchorage Data: Are They Representative?This paper presents the results of a study designed to assess the representativeness of realized samples of recent arrestees selected for the Arrestee Drug Abuse Monitoring (ADAM) program in Anchorage, Alaska. Because one of the most important goals of the ADAM program is to produce scientific information on the prevalence of alcohol and drug use behaviors among arrestees that is generalizable to an entire local arrestee population, establishing the representativeness of realized samples (or isolating inherent biases) is an essential first step to meaningful use of these data to address locally defined problems. In order to determine the reasonableness of inferences grounded in realized samples of ADAM respondents, an analysis was done comparing various characteristics between each stage of the sample selection process including the census of eligible arrestee population, the designed ADAM arrestee sample, arrestees available for interview, arrestees actually interviewed (“realized” sample), and arrestees that provided urine sample (“realized” sample). If the realized samples are similar to the census we can have a greater degree of confidence in our capacity to describe the population of Anchorage arrestees using ADAM data. Also, if it happens that departures are detected between realized samples and the arrestee census we are better positioned to condition the inferences made by integrating these discerned biases into our conclusions.
Anchorage Community Survey 2007 Survey Sampling Design: Power and Sample SizeThis working paper documents the power analysis, literature review, and precision considerations contemplated in designing the Anchorage Community Survey’s (ACS) 2007 sampling design. The ACS will obtain at least 30 completed surveys from individuals in each of the 55 census tracts that make up the Anchorage Municipality, allowing us to discern a fairly small effect size of 0.30 with our smallest anticipated intraclass correlation and a moderate effect size of 0.40 with our largest anticipated intraclass correlation, both at 0.80 power level. This cluster sample size and number of clusters should yield sufficient precision to allow good estimation of variance components and standard errors, acceptable reliability estimates, and reasonable aggregated measures of constructed neighborhood variables from individual survey item responses.
Converging Science and Practice in Analyzing Evaluation DataA strategy is presented for converging science and practice which focuses on the needs of scientists and policymakers in analyzing evaluation data. Emphasis is placed on employing powerful statistical techniques that maximize the evaluators' confidence in their results. Attention is also drawn to the need for producing results which can be easily communicated to and interpreted by policymakers. In regard to these requirements, the discussion concerns application of four statistical techniques: factor analysis, Guttman scalogram analysis, multiple classification analysis and cross-break analysis. Each statistical analysis technique is described as to its value in evaluation research for dealing with problems known to inhibit the convergence of science and practice. The application of these techniques is demonstrated by illustrations taken from previous evaluation studies. The paper concludes with implications for stimulating the extent and quality of evaluation use.
Introduction to Data CollectionThis Powerpoint presentation illustrates the fundamentals of data collection through the example of an evaluation of Teens Acting Against Violence (TAAV), a violence prevention and youth empowerment program for teenagers operated by the Tundra Women’s Coalition in Bethel, Alaska. Key results from the evaluation are presented.