EXTENDED USES OF LINEARIZED NONLINEAR REGRESSION FOR RANDOM-NATURE SIMULATIONS.
Themis Signal Analysis Statistics Research Program,
SOUTHERN METHODIST UNIV DALLAS TEX DEPT OF STATISTICS
Pagination or Media Count:
Linearized nonlinear regression introduced in ref. 1 has substantial curve-fitting capability, computational simplicity, ability to isolate and investigate effects of interest, etc. A probability model was developed that yields approximate median estimates and confidence intervals for the individual regression coefficients. This model is applicable for random-nature simulations if the simulations are statistically independent. This approach allows the outcomes for wide classes of combinations of values for simulation inputs that specify the situation simulated to be estimated from a moderate number of simulations. In this extension of the method, the probability model is slightly changed and approximate results with greater practical utility are developed. Median estimates, confidence intervals, and significance tests are developed for specified linear functions of the regression coefficients that are associated with the simulation inputs. Also, properties of least-squares estimates for specified linear functions of the regression coefficients are examined. Author
- Statistics and Probability