H518 Statistics in Health Administration
Final Assignment
Due May 3rd, 2010
Your third assignment is to perform a regression analysis to test for relationships between variables. Specifically, you should
1. Identify a question of interest to a hypothetical decision maker
2. Perform an appropriate regression analysis that addresses this question
3. Include the statistical results and their implications for the decision maker in a short manuscript.
You may analyze the same issues you did in previous assignments, but this is not required.
Data to evaluate your hypothesis must come from one of the following data sets:
1. the BRFSS 2008 (http://www.cdc.gov/brfss/index.htm)
2. the CTS Household Survey 2007 (http://www.hschange.org/index.cgi?data=02)
3. the CTS Physician Survey 2004-5 (http://www.hschange.org/index.cgi?data=04)
These data are available under Q:\aholmes\H518 in BS1000 (along with codebooks).
Choice of type of regression analysis should reflect the nature of your dependent variable. The regression analysis must have at least two independent variables (an “intervention” variable and at least one control). The manuscript by Coggon et al. (http://www.bmj.com/epidem/epid.html) may help you identify potential confounding variables. Specification may also be motivated by one of the models provided on the class website (the Evans’ model of health status, the Anderson’s model of health care use, or O’Neill and Kuder’s model of physician practice. Where warranted, appropriate inclusion of dummy variables and non-linear and interaction terms will also be expected.
Results should be written up in the form of a short manuscript (e.g., 2,000 – 2,500 words, exclusive of abstract, figures, tables, and references; no more than 4 tables/figures). Manuscripts should have the same basic structure as the structured abstract previously completed, but with each section fully developed (you don’t have to follow the Uniform Requirements for Biomedical Journals, but you may find the instructions for manuscript preparation helpful: http://www.icmje.org/manuscript_1prepare.html). Empirical results should be presented in tables/figures which may be either embedded in the paper or attached as appendices. Follow Miller’s textbook to present regression analyses effectively. All sources must be properly cited.
You will be assessed, in part, according to how well motivated the analysis is (does it address a problem that is important), the appropriateness and sophistication of the statistical analysis, and the effectiveness with which the results are presented.
Manuscripts should be submitted electronically (to ) by May 3rd, 2010. In addition, include the final regression output(s) on which the paper is based, and tables of summary statistics for the dependent variable and each independent variable used in the final regression equation
Papers must be based on original work undertaken exclusively for this class and be completed independently.