Microsoft word - finalassignment2010.doc

Final assignment - 2010
Statistics part II – Regression Analysis and SPSS
In a cross-sectional study, the relationship between obesity and estrogen levels was studied for 211 pre-menopausal women (161 African-American and 60 Caucasian). In particular, one type of estrogen, the serum estradiol level is studied, since it is a strong risk factor for breast cancer. Potential confounders for the relationship between obesity and serum estradiol are ethnicity, age, any children, age at first birth, and age at menarche. As a definition of obesity, you may use either the variable body mass index (BMI, kg/m2)
and/or waist-hip-ratio (WHR). These variables are measured on a continuous scale, but you may also choose to use a cut-off value to define obesity as long as you give an explanation why this cut-off value was chosen (use references). The same applies to the outcome variable, i.e. estradiol level can be analyzed as a continuous variable or as a binary variable using a
chosen cut-off value. Note that these choices influence the analysis method that should be The data are given in finalassignment2010.sav. A description of the variables is given
Variable Description
Ethnicity (1 = African-American; 0 = Caucasian) Age at first birth (years; coded as 0 for no children) Analyze these data and report the used method, modelling strategy (if applicable) and results of the analysis as if it was part of a scientific paper1. More specifically, describe the analysis in three sections: “methods”, “results” and “conclusions”. In the “methods” section,
the variables and analysis methods that were used have to be described in a way that the reader (with access to the data) can apply these methods to the data and obtain the same results. The same applies to the approach you used to assess confounding and effect- modification, and to the modelling strategy that was used. In the “results” section, the
outcome of the analyses should be reported, where only the main results are discussed in detail and the complete results are summarized in Tables and/or Figures. Make sure that no conclusions are included in this section, since they go into the “conclusions” section.
The report should at least (directly or indirectly) answer the following questions: 1. Does the effect of obesity on serum estradiol depend on ethnicity? 2. Which aforementioned variables are confounders and/or significant risk factors? (do not forget to give the definition of confounding that you used) 3. Are the assumptions of the used method met? (also check for (multi)collinearity and influential outliers) 4. Is the effect of obesity significant? Present the crude and adjusted results, including the effect size (e.g. odds ratio), p-values and confidence intervals of the effects of (paper versions or by email (ask for confirmation)):
Deadline:
Wednesday 25th of August 2010
Bjorn Winkens Methodology and Statistics P.O. Box 616 6200 MD Maastricht 1 This means that you should write the report in English and that the report does not include SPSS output tables (figures are allowed). If you have any questions or problems regarding the assignment, please contact me by phone (043-3882433) or by email. I am unavailable between 17th of July and 15th of August. If you need more time, please contact me as soon as possible, but at least before the 25th of August!!!

Source: http://www.stat.unimaas.nl/bjorn/statisticspart2/2010/finalassignment2010.pdf