Activity Number:

486
 Computing Kaleidoscope

Type:

Contributed

Date/Time:

Wednesday, August 1, 2018 : 8:30 AM to 10:12 AM

Sponsor:

Section on Statistical Computing

Abstract #329895

Presentation

Title:

Nearly Best Confidence Intervals

Author(s):

George Terrell*

Companies:

VA Poly. Inst. & State Univ.

Keywords:

Confidence Intervals;
Estimation;
SmallSample Asymptotics

Abstract:

A confidence interval for a parameter theta is a range of values determined by experimental information that has a certain 1 minus alpha probability of covering the true value of that parameter. The coverage probability determines only one of the limits of an interval. The best way to determine the other limit is the condition Likelihood(theta) > = c(alpha). In unsymmetric families, this has no simple formula and must usually be found numerically. We here propose closedform approximations to the limits in several important families.
