Variable selection for propensity score model.

 
   Lead Investigator:    Dingke Tang
   Institution      :    University of Toronto
   E-Mail           :    tdk@mail.ustc.edu.cn
   Proposal ID      :    1302



   Proposal Description:
   Hi I am a student of University of Science and Technology of China (USTC), and now I am a visiting student of University of Toronto.     In propensity score variable selection, it is well know to researcher that the optimal procedure is first select predictor of outcome and exclude predictor of treatment. Based on this argument we developed a powerful method to help us do variable selection, and in simulation study, our procedure is much powerful than other existing method. Now we need a real data to check the power of our method. This data should be high dimensional since we want to check it's performance in high-dimensional setting. We found that NACC data is perfect match to what we want. For it's high dimensional and there have a lot of literature to verify if we can successfully identify true predictor of outcome. we can put one co-variate as treatment and one as outcome, the rest variables could be regarded as possible predict variables. I hope that you can give us the access of your data, thank you very much for reading my massage.    We need NACC data to check if our proposed data mining method work. We            developed an algorithm to select predictor of outcome, and it can work in high dimensional setting where p>>n. So we found NACC data is perfect for us. We mainly interested in AD data. And we want to longitudinal data.