Abstract:
|
Batched data latent arose from many practical problems. This type of data has the feature that it can be divided into a member of batches, each batch is a linear combination of certain number of latent responses, only the total response of each batch is observed. Such data are non-iid, with unknown dependence structure within each batch. Some existing methods, such as least squares with/without constraint(s), do not give satisfactory results. Nonparametric models are difficult to apply for this type of data. Here we explore the Gaussian model for such data. Simulation studies are performed to evaluate the proposed method.
|