Recent Work
Thesis Assistance
Industry: Education
Overview: Thesis project determining whether the views of breastfeeding newborns are different between expectant mothers who smoked versus those who were nonsmokers.
Solution: Survey based study in which data was screened for accuracy, reliability, and consent. Data was entered into Excel before being imported into SPSS. Normality tests were run to determine proper tests and reliability analysis was conducted to see the efficacy of the survey. Following, Wilcoxon signed-rank tests were used to determine differences in the distribution of the measurements.
Methodology: normal probability plots, reliability analysis, Wilcoxon signed-rank, Spearman’s rank correlation test
Dissertation Assistance
Industry: Education
Overview: Doctoral dissertation determining what factors could contribute to varying degree of attendance at Division I men’s soccer games.
Solution: Survey based study going out to all Division I collegiate soccer programs. Data was entered into Excel and screened for outliers and erroneous errors. Free text was grouped into different categories based on response. Data was imported into SAS and normal screening tools were used, such as frequency counts, descriptive statistics to look for more outliers that initial screening missed. Normal probability plots were used to determine the type of tests that could be used. Analysis included ANOVA and different types of regression modeling, which concluded with backwards stepwise linear regression as the best model. Minitab was used briefly as a crosscheck.
Methodology: normal probability plots, ANOVA, linear regression, logistic regression, backwards stepwise linear regression
SPSS Training
Industry: Education
Overview: Training requested by the New York State Department of Child and Family Services
Solution: Prepared a customized training program for beginning and intermediate SPSS users. Delivered and executed lessons focused to their specific needs.
Methodology: SPSS
SAS Training
Industry: Education
Overview: Training requested by the New York State Department of Health
Solution: Prepared a customized training program for beginning and intermediate SAS users. Delivered and executed lessons focused to their specific needs.
Methodology: SAS
Research
Industry: Health Care
Overview: Project scope to try to predict length of stay for a newly admitted mental health patient based on specific inpatient forms filled out by the doctor at time of admission.
Solution: Obtained and screened data. Created predictive models to obtain good estimate of levels of care based on several independent variables.
Methodology: Linear regression, logistic regression, ANOVA, MANOVA, factor analysis, transformations
Excel Training
Industry: Education
Overview: Training requested to develop skills in statistical analysis in Excel, including t-tests and chi-square.
Solution: Prepared a customized training program for beginning and intermediate SAS users. Delivered and executed lessons focused to their specific needs.
Methodology: Excel
Validation of New Software Package
Industry: Health Care
Overview: Request to test the validity and reliability of a new system that tracks authorizations and reviews for a national health care client.
Solution: The major variables that were used in the project consisted of those related to the risk/impairment ratings that are completed by providers. In addition to evaluating the reliability of the assessment tool, other analyses focused on how different variables loaded on key factors. Correlations among the variables were also examined to determine if certain ones were deemed not related to the others in such a way that the variable becomes nonessential in further analysis
Methodology: factor analysis, reliability analysis, correlation analysis
Research
Industry: Health Care
Overview: Upon brief summarization of the HEDIS ambulatory measure, it was viewed upon that certain high volume facilities had a better percentage of patients who followed up their inpatient discharge within 7 and 30 days with an outpatient level of care then other facilities. It was desired to determine why this might occur and several hypotheses were considered.
Solution: The first analysis that was run was to determine whether or not the hypothesis was correct in the assumption that different facilities did indeed have different follow-up proportions when compared to other facilities. Chi-square tests were used in determining that both measures of 7 and 30 day had statistically significant difference in proportions when compared across facilities. This was the initial test to ensure that the visualization of the data did produce statistically significant results and thus, can continue on in attempting to determine the reason(s) behind the differences.
Methodology: Chi-square, proportion analysis, Marascuilo procedure
Research
Industry: Health Care
Overview: Intensive case management (ICM) is a form of case management aimed towards the population with the greatest service needs. Examples of the types of individuals who could benefit from ICM programs include those with a history of multiple hospitalizations for mental illness, a serious co-morbid physical ailment (i.e. cancer), eating disorders, schizophrenia and other psychotic disorders, substance abusers, and those with a mental illness who have experienced problems with the law.
The study describes the ICM program, including the demographic characteristics of members in the program and its effectiveness. The primary outcome indicators examined whether there were significant decreases in cost and/or service utilization of members subsequent to enrollment in the ICM program.
Solution: In addition to basic descriptive analysis to provide a general overview of the program, several paired analysis were conducted on claim utilization and cost, both prior to and post discharge from the program. The inferential analyses only included members who were successfully discharged from the ICM program with a sufficient claim lag (3 months) for an appropriate follow-up period.
Methodology: two-sample t-test
Research
Industry: Health Care
Overview: A study to determine the effect of diabetes and co-morbid depression had a negative effect on health care costs and utilization. Those members selected were then split into two groups: 1) members with co-morbid depression; 2) members without co-morbid depression. These two cohorts were analyzed to determine if members with co-morbid depression had an increased likelihood of inpatient admission, higher future costs, higher risk scores, and other indicators of negative outcomes.
Solution: Data was pulled and screened for outliers before being put through a string of inferential tests for cost and utilization comparison between the two cohorts. Risk was also assessed to determine differences.
Methodology: two-sample t-test, proportion analysis, relative risk calculation