1. The investigators did not adjust for childhood socioeconomic status (SES) in their analysis, which you suspect may be a potential confounder. You contact the investigators, who send you the original dataset, and you observe the following 2x2 tables when you stratify the data on SES. Calculate the odds ratios stratified by SES. In comparing the crude and stratified ORs, what do the odds ratios suggest? Is there confounding by SES? (5 points)

High SES:

Candy Consumption	Violent Behavior	No Violent Behavior
Daily	3	988
Less than Daily	4	1365

Low SES:

Candy Consumption	Violent Behavior	No Violent Behavior
Daily	15	2643
Less than Daily	11	1913

2. List two ways that the investigators could have controlled for childhood SES in the design phase of the study. (2 points)

3. The investigators looked at several potential confounders in this study, including: sex, education, childhood behavior & aggression, mental ability at 5 years, parenting styles, late birth, health visitor screening, maternal circumstances in childhood, access to a car in adulthood, and living in a rural area in adulthood. (4 points)

State two ways that residual confounding may have occurred, using specific examples from the study. Residual confounding refers to bias remaining within the estimated associations (i.e. RR, OR, etc.) due to not accounting for all true confounders of the exposure and outcome of interest within study design or analysis (i.e. confounders that were not controlled for).