Hello, everyone. I am actually quite new with Statalist, and just beginning to learn Stata beyond what was taught in our syllabi. I need your help with IV estimation. I wish to estimate mortality risk with BMI as main predictor (with survival analysis). To address the issue of reverse causation involving BMI and comorbid illness, I would like to use BMItime–1 as instrumental variable for BMI... with both BMIs as ordinal variables (underweight, normal[baseline], overweight, obese 1, obese 2, obese 3). What should I use in Stata? Is it ivregress or ivpoisson. Also, can anyone help me on how to code this in Stata? So far, the Stata manual hasn't been really helpful (the instrument is treated as continuous), and I've searched as extensively as I can, but came up with nothing. Do I create a separate dummy variable for each BMI class (e.g., BMI_time-1_underweight = 0 or 1, etc.)?

For completion, my other exogenous variables are age, sex, and current smoking status, and my other instruments are diabetestime-1 (0 or 1) and cardiovascular diseasetime-1 (0 or 1) and smoking status at time-1 (smoker vs nonsmoker).

Thank you all so much for your time and understanding.