Policy Research

As a Human Services Research Assistant/Programmer at Mathematica Policy Research, Inc. in Princeton, New Jersey, Ravaris made significant statistical programming contributions to the following projects:

Social and Character Development Research Program (SACD). Programmed SACD Growth Curve Analysis.  SACD implements and evaluates promising school-based interventions designed to promote positive social and character development among elementary school children. The Growth Curve analysis uses five rounds of data to test for impacts of the treatment over time in a hierarchical linear model with sample weights and clusters at the person and school levels. Additional analyses tested for differences in impacts based on gender, original cohort/new-entrant status, and measures of socioeconomic risk, family risk, and community risk and child behavior risk scales.

Building Strong Families (BSF). Analyzed implementation data to assess the extent to which couples randomly assigned to treatment received key components of the intervention. This helps assess the benefits and costs of different implementation practices. Contracted by DHHS, BSF is a large-scale demonstration and evaluation of interventions designed to enhance child and family well-being by helping unmarried parents strengthen their relationship through family counseling. Treatment includes attending group sessions where trained facilitators lead couples through one of three curricula designed to strengthen relationship and parenting skills.

National Evaluation of Reading Comprehension Interventions. Created an organized system of programs and directories for cleaning and analyzing data, and organizing output. Ran impact analyses, identified data inconsistencies, and produced clean data and documentation for external use. This experimental study for the US Department of Education evaluates the impact of four interventions in social studies and science on fifth-grade reading achievement.  Eighty-nine schools were randomly assigned to one of four treatment groups, or to the control group. After baseline testing and a period of exposure to the interventions, we analyze differences in follow-up measures of reading comprehension ability in a hierarchical linear model with school level clustering, non-response weights, and multiple comparison adjustments.

Precision Gains from Publically Available School Proficiency Measures Compared to Study-Collected Test Scores in Education Cluster-Randomized Trials. Education randomized control trials (RCT’s) usually issue baseline tests as a benchmark for identifying possible difference between treatment and control groups. This is reliable, but costly to implement. This research investigates the possibility of using less costly school level baseline covariates instead. We investigate the possibility of designing less costly studies that use publicly available school level covariates to achieve the same minimal detectable effects. Programming contributions include downloading and organizing over 500 school level datasets; writing two SAS programs with the combined flexibility needed to clean and reshape hundreds of datasets with varying file structures; merging the data with original study data to run various analysis and create tables  using R, SAS, and excel.

Kaufman Campus Initiative. Cleaned and analyzed survey data about college student’s perceptions of entrepreneurship. Research assessed the use and impacts of grants awarded to eight universities that pledged to make entrepreneurship education available across campus, with the goal of transforming the way that entrepreneurship is viewed, taught, and experienced.