Weighted-average least squares (WALS): Confidence and prediction intervals

We extend the results of De Luca et al. (2021) to inference for linear regression models based on weighted-average least squares (WALS), a frequentist model averaging approach with Bayesian flavor. We concentrate about single focus parameter, interpreted as causal effect policy or intervention, in presence potentially large number auxiliary parameters representing nuisance component model. In our Monte Carlo simulations we compare performance WALS that several competing estimators, including unrestricted least-squares estimator (with all regressors) and restricted no regressors), two post-selection estimators alternative selection criteria (the Akaike information criteria), various versions (Mallows jackknife), one version popular shrinkage adaptive LASSO). discuss confidence intervals parameter prediction outcome interest, conclude leads superior intervals, but only if apply bias correction.