This PR aims to change the way superblock is handled to retrieve classical methods of the literature. After this PR,
when superblock = TRUE, deflation is performed on the superblock, and regular blocks are retrieved from the deflated superblock;
as a consequence of the previous point, astar is no longer meaningful for the regular blocks, so when a superblock is used, astar is not a list but a matrix associated with the astar of the superblock;
scale_block behavior changed to enable a new option: scale_block = "lambda1". This option scales each block by the first eigenvalue of its empirical covariance matrix. scale_block = "inertia" or scale_block = TRUE gives the same result as before (i.e., the block is divided by the sum of its empirical covariance matrix's eigenvalues). Any other value of scale_block leaves the block unchanged.
Tests to compare RGCCA with other methods from the literature have been updated in test_rgcca.R.
This PR aims to change the way superblock is handled to retrieve classical methods of the literature. After this PR,
superblock = TRUE, deflation is performed on the superblock, and regular blocks are retrieved from the deflated superblock;astaris no longer meaningful for the regular blocks, so when a superblock is used,astaris not a list but a matrix associated with theastarof the superblock;scale_blockbehavior changed to enable a new option:scale_block = "lambda1". This option scales each block by the first eigenvalue of its empirical covariance matrix.scale_block = "inertia"orscale_block = TRUEgives the same result as before (i.e., the block is divided by the sum of its empirical covariance matrix's eigenvalues). Any other value ofscale_blockleaves the block unchanged.Tests to compare RGCCA with other methods from the literature have been updated in
test_rgcca.R.