. Author manuscript; available in PMC: 2020 Apr 17.

Published in final edited form as: Ann Appl Stat. 2018 Jul 28;12(2):940–970. doi: 10.1214/18-aoas1175

Table 3:

Evaluation of correlation matrix estimates from simulation study. “LPoC” refers to our estimator, which uses Laplace priors on correlations. MAE is mean absolute error. MSE is mean squared error. Averages over “all elements” exclude diagonal elements, which are fixed at zero by all methods. The lowest (best) values are shown in bold.

Values of ε_t estimated with MCMC
	Estimator	MAE	MSE
All elements	Pearson Ledoit-Wolf LPoC	0.253 0.193 0.090	0.098 0.055 0.028

True correlation = 0	Pearson Ledoit-Wolf LPoC	0.270 0.190 0.049	0.109 0.053 0.012

True correlation = 0.5	Pearson Ledoit-Wolf LPoC	0.201 0.200 0.214	0.066 0.060 0.074

True values of ε_t used
	Estimator	MAE	MSE

All elements	Pearson Ledoit-Wolf LPoC	0.227 0.182 0.078	0.079 0.047 0.022

True correlation = 0	Pearson Ledoit-Wolf LPoC	0.244 0.162 0.041	0.089 0.039 0.010

True correlation = 0.5	Pearson Ledoit-Wolf LPoC	0.176 0.243 0.190	0.051 0.073 0.058