COMPSTAT 2022: Start Registration
View Submission - COMPSTAT2022
Title: Constructing likelihood-ratio-based confidence intervals for multiple threshold parameters Authors:  Luiggi Donayre - University of Minnesota - Duluth (United States) [presenting]
Abstract: A procedure to compute sequential (one-at-a-time) asymptotically-valid confidence intervals for $n$ threshold parameters is proposed. Because the sequential procedure yields T-consistent estimates of the threshold parameters, the asymptotic distribution is identical to that for a model with multiple thresholds simultaneously estimated. Consequently, the limiting distribution of the likelihood ratio statistic, conditional on $(n-1)$ consistently estimated threshold parameters, is asymptotically free of nuisance parameters and critical values to construct confidence intervals for multiple threshold parameters can be derived. Using Monte Carlo simulations, conservative likelihood-ratio-based confidence intervals are shown to exhibit coverage rates at least as high as the nominal level for multiple threshold parameters, while only including relatively few observations of the threshold variable for each confidence interval.