Confounding and Effect Modification: Directed Acyclic Graphs and Interpreting Interaction - Assignme

Assignment: Confounding and Effect Modification

To complete this section, please review the articles on DAGs in the Week 5 Learning Resources. You may use any drawing tools available to you to create the DAGs requested below, although it does not need to be sophisticated software; for example, Microsoft Word’s Insert Shapes and Text Box features are sufficient tools to create these diagrams.

For each of the following epidemiological scenarios, draw a DAG that represents the relationships between the variables: (5 points each)

• A researcher is studying the association between family income (I), family history of diabetes (H) and the occurrence of diabetes (D) as an adult. Income and family history are believed to be independent risk factors for diabetes.

• A researcher is studying the association between human papilloma virus (V) and the incidence of oral cancer (C). Smoking (S) is related to both variables. Alcohol use (A) is an independent risk factor for oral cancer.

A researcher studied women with breast cancer to better understand the association between self-rated health and all-cause mortality (Prehn, 1996). Women were interviewed twice, and survival models were used to assess the association between self-rated health 12 months after breast cancer diagnosis and all-cause mortality. The following results were obtained from three different models:

• State in words the meaning of the odds ratios for self-rated health in model 2. Be sure to include information on statistical significance. (5 points)

• State in words the meaning of the odds ratios for self-rated health in model 3. Discuss what these models show about the role of confounding in the self-rated health/mortality association. Justify your response and state your overall conclusions based on the analyses. (15 points)

Assessing and Interpreting Interaction/Effect Modification (15 points)

• Fill in the table below for columns A and B. The table shows different associations where Factors X and Z are being measured for their effect on an outcome. For each effect measure there are shown three results: a result with Factor X only (X+Z-), a result with Factor Z only (X-Z+), and the Observed Joint Effect for both factors (X+Z+). In column A, indicate whether the interaction is positive or negative, and in column B, indicate whether the interaction is additive or multiplicative. If no interaction is present, put “None” in columns A and B. (10 points)

• Middleton and colleagues (2010) conducted a study to look at the relationship of physical activity over the life course (i.e., at different ages) and the development of late-life cognitive impairment. They analyzed cross-sectional data using logistic regression, and reported the following results:

Describe the results of the study as represented in the table above. Be sure to address main effects of the different subgroups as well as any evidence of effect modification. Don’t forget to address the statistical significance of the findings. (5 points)