Soft maximin estimation for heterogeneous data

Extracting a common robust signal from data divided into heterogeneous groups is challenging when each group - in addition to the contains large, unique variation components. Previously, maximin estimation was proposed as method presence of noise. We propose soft computationally attractive alternative aimed at striking balance between pooled and (hard) estimation. The provides range estimators, controlled by parameter ζ > 0,, that interpolates least squares By establishing relevant theoretical properties we argue statistically sensibel attractive. demonstrate, on real simulated data, can offer improvements over both OLS hard terms predictive performance computational complexity. A time memory efficient implementation provided R package SMME available CRAN.