Bayesian regression for large dataset/matrix

Question

I have encountered issues of fitting BKMR in large dataset. A BKMR model was constructed with binomial link.

model_bkmr <- kmbayes(y = Y, Z = Z, X = X,id=group, iter = 10000,family = "binomial",
                      verbose = FALSE, varsel = TRUE)

of which y is an vector of 600000 numbers (0/1), X is a matrix of numeric predictors with 6 cols and 600000 rows, X is a numeric covariate matrix with 7 cols and 600000 rows, group is an vector of 600000 numbers (1-6) indicating six groups for the 600000 rows. The model can not be fitted due to the expensive matrix computations, is there anyway to cope with such situation?

Answer 1

I found that through the way of partition the data into independent subsets and compute posterior for each subset, it is then possible to recover the full posterior (Sudipto Banerjee: High-dimensional Bayesian geostatistics).