I work at a local health department evaluating some of our work in obesity prevention. We are interested in looking at stress, social support and physical activity on obesity and waist circumference (WC). Our sister site only collects BMI, while we collect both BMI and WC. We are interested in estimating a “calibration factor” for our other site so we can compare the effects on waist circumference.

To do this we did a split sample estimation, dividing half of our sample to estimate the correction factor, and validated the effects using the other half. Next we’d like to try bootleg simulation methods but not sure how to code the simulation in STATA. We are assuming a linear relationship between stress and BMI.

Our model:
X= stress
Y=Waist Circumference (the Y we have but our other site doesn’t)
Y*= BMI (Like to estimate the effects of x on y when you only have BMI)
C= set of confounders

Thank you!