Abstract:
|
In principle, it can be challenging to integrate data measured on same individuals occurring from different experiments and model it together to get a larger understanding of the problem. Canonical Correlation Analysis (CCA) provides a useful tool for establishing relationships between such data sets. When dealing with high dimensional datasets, Structured Sparse CCA (sSCCA) is a rapidly developing methodological area which helps to extract signal from vast amount of noise present in the data considering its structure which results in sparse direction vectors which are used to calculate CCA. There is less development in Bayesian methodology in this area. In our project we use a latent variable model, whereby using horseshoe prior, we bring in sparsity at projection matrix level, as well as at the structure level by using graphical horseshoe prior on covariance matrix to model datasets on same individuals from two experiments. We compare our results with some competing methods in a series of simulation studies.
|