Meta-analysis of Disjoint Sets of Attributes in Large Cohort Studies

We will introduce the problem of classification in large cohort studies containing heterogeneous data. The data in a cohort study comes in separate groups, which can be turned on or off. Each group consists of data coming from one specific measurement instrument. We provide a “cross-sectional” investigation on this data to see the relative power of the different groups. We also propose a way of improving on the classification performance in individual cohort studies using other cohort studies by using an intuitive workflow approach.