AShrink: Adaptive shrinkage for effect estimates and regression in Python

Effect estimates often differ in precision. Adaptive shrinkage uses information shared across estimates to learn a prior distribution and compute posterior summaries. AShrink provides this approach in Python for supplied effect estimates and for selected regression models. It supports Gaussian and Student-t measurement errors, conjugate Poisson and binomial models, multivariate Gaussian effects with supplied covariance patterns, and regression with adaptive normal-mixture priors. We distinguish conditional posterior calculations from variational approximations used in regression. Fourteen fixed cases agree with results exported from installed R reference packages under aligned settings. Installed wheels pass 116 tests in each of two macOS environments, two Linux environments and a native Windows environment. These checks support numerical agreement in the specified cases; they do not establish repeated-dataset interval coverage or comparative predictive performance. AShrink includes distribution methods, sign and null probabilities, prediction, tables, plots, and an editable reference manual.