Jackknifing estimated weighted least squares: Jewls
Jack P. C. Kleijnen, P. Karremans, W. Oortwijn, W.J.H. van Groenendaal · Communication in Statistics- Theory and Methods · 1987
This paper investigates regression analysis of experimental designs with replications, assuming variance heterogeneity, possibly combined wioth nonnormality. These replication yield variance estimators which result in Estimated Weighted Least Squares (EWLS). Jackknifing yields confidence intervals for these nonlinear EWLS estimators. The Validity of these confidence intervals is examined in a Monte Carlo experiment. Jackknifed EWLS estimators result in better confidence iintervals than simple EWLS.