Stochastic approximation models in estimating productivity

From MaRDI portal





A new stopping rule for stochastic approximation is presented and properties are investigated in a novel application of the general calibration problem when underlying assumptions are not satisfied. Of particular interest is presentation of effects of regression models under model misspecification. Generally, stochastic approximation methods offer useful sequential alternatives to fixed sample methods of numerical analysis and regression in estimating a fixed point of a function. Specifically, early production data from oil and gas leases is used to estimate the time to depletion of recoverable reserves. The performance of stochastic approximation is evaluated and the performance of traditional methods of regression and the controversial method reverse regression is illustrated. Recommendations are made which utilize multiple estimation criteria of economy of sample size, precision, and accuracy.











This page was built for publication: Stochastic approximation models in estimating productivity

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3694510)