get_posterior_FUN.RdThis function will apply the supplied function to each posterior sample of the chain. Variables referenced in FUN will be selected from the following search locations (in this order): 1) Posterior$posteriorSample_params 2) data_matrices 3) priors 4) Posterior$posteriorMean_params 5) current_state 6) calling environment (ex. sapply) 7) global environment
get_posterior_FUN(MegaLMM_state, FUN, samples = NULL, mc.cores = 1)A MegaLMM_state object including a re-loaded Posterior list
Operations to be applied to each posterior sample. Write as if this were operating within current_state. Can use priors, data_matrices, and other elements of current_state
(optional) vector of sample indexes to use in the computation
(optional) number of cores to use for computations. See note about memory requirements.
array of n_samples x dim1 x dim2 where dim1 and dim2 are the dimensions of the calculated parameter per posterior sample
If mc.cores > 1, the operation will be parallelized. To reduce memory requirements, if mc.cores > 1, a PSOCK cluster is created. This cluster only copyies in data when it is actually used. The code is written so that only necessary objects are used by this cluster, not the whole user's environment. Inside this cluster, mclapply is run, which forks the process inside the PSOCK cluster. Creating the cluster takes some time, so often mc.cores=1 is faster. If the operation is large and Posterior has many samples, this can still create memory issues, so limiting mc.cores might still be necessary