Change of support (spatial downscaling/upscaling)

[AlePalu]

In applications it might be necessary to consider responses and/or covariates defined on different spatial grids and/or different areal partitionings, e.g. we have a covariate defined on a regular grid, while another is available only at areal level, while keeping the same underlying domain discretization.

Other than evaluating the changings in the estimation functional, for sure, different covariate/responses are required to be attached with different spatial informations. Not only this, the sampling design cannot be anymore a global property of a model, as, in this case, it would be a data-dependent information (e.g., is nonsense to define the sampling property of a model when the model is constructed).

As a result of this, it would be interesting to question about a library support for multiple domains (e.g. responses and covariates have potentially different domains, each with its own discretization).