LD shrinkage estimation in GCTB

[xiangzhu]

It seems that GCTB software has implemented Wen-Stephens shrinkage LD estimator (see also https://github.com/stephenslab/rss-private/issues/64):

Shrunk LD matrix
The shrinkage estimator for the LD correlation matrix was originally proposed by Wen and Stephens (2010). The estimator shrinks the off-diagonal entries of the sample LD matrix toward zero. Zhu and Stephens (2017) used the shrinkage estimator in their Regression with Summary Statistics (RSS) methodology and showed empirically that it can provide improved inference. The shrinkage estimator overcomes some of the issues arising from approximating the full LD matrix in the summary-data based model using a subset of the full LD matrix and constructing the matrix from a reference. The GCTB implementation is a C++ port from that provided with the RSS software and has been adapted for use with the GCTB software.

GCTB software doc page: http://cnsgenomics.com/software/gctb/#SummaryBayesianAlphabet