In the next model, we will add a set of possible confounding variables to the regression: sex, parents’ education and family social class. First, we will add sex where 0=Male and 1=Female. As explained in previous sections, this type of binary variable is also known as a dummy variable. In our analysis, the reference group will be ‘male’ (as this group is coded as 0). We are also going to include a few family background factors in the model: whether the cohort’s mother (n016nmed) and father (n716dade) left school at the minimum age or not, and the social class of the study participant’s father (n1171_2). Social class n1171_2 has 5 categories: ‘I/II Prof & Managerial’, ‘III Skilled non-manual’, ‘III Skilled manual’, ‘IV Partly skilled’ and ‘V unskilled’. With multi-category variables such as this, you can use the prefix of ‘i.’ in the variable name i.n1171_2 and Stata will automatically create dummy variable(s) for each category. The first category ‘I/II Prof & Managerial’ will be treated as the reference category.
Interestingly in the output we can see that ‘general ability’ is significant in the ‘obese’ versus ‘normal/healthy’ BMI comparison, but not in the ‘overweight’ versus ‘normal/healthy’ BMI comparison after controlling for all the other predictors. A 1 unit decrease in ‘general ability’ test score is associated with a .014 increase in the relative log odds of being obese v normal/healthy BMI at age 42. Father’s social class also predicts obesity; it is associated with the odds of the study participant being overweight compared to normal/healthy BMI in the study participant. Males (compared to females) and participants whose mothers left education at the minimum age were more likely to be overweight or obese compared to normal/healthy BMI.
For our final model, we are going to include bmi11, the BMI of the participant when they were aged 11. Doing so means that we will be adjusting for participant’s baseline BMI, and that will allow us to focus on the subsequent change in BMI from age 11 to age 42, and therefore to measure both BMI and general ability over a comparable period, from childhood to middle age.
In the output above, we can see that after controlling for BMI at age 11 ‘general ability’ is significant in the comparison of obese versus normal/healthy BMI, but not in the overweight versus normal/healthy BMI comparison. A 1 unit decrease in ‘general ability’ test score is associated with a .017 increase in the relative log odds of being obese versus normal/healthy BMI at age 42. Lower parental social class, compared to professional and managerial is also important. In addition, as in the previous model, males are more likely than females to be either overweight or obese than to have a normal/healthy BMI.
The Learning Hub is a resource for students and educators
tel | +44 (0)20 7331 5102 |
---|---|
closer@ucl.ac.uk |
Sign up for our email newsletters to get the latest from CLOSER
Sign up