Repeated Significance Tests for Exponential Families.
Abstract
This report considered the significance levels and powers of repeates significance test (RTS). Typically these quantities cannot be calculated exactly and some sort of approximations are required. Satisfactory approximations of significance levels of RST for exponential families have been obtained by Lalley (1983) and Woodroofe (1978). In this report another method due to Siegmund (1985) in the special case of normal observations with known variance is developed. The main advantages that are claimed for this method are two-fold. First, it can be used to approximate the power of the RST. Second, it enables one to estimate both powers and significance levels of the modified RST. The approximations are also useful in determining confidence intervals. The proof of a nonlinear renewas theorem for conditional random walks and some numerical results are also given. Additional keywords: Distribution functions; Tables(data); Random walks; Null hypothesis case. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1985
- Accession Number
- ADA159547
Entities
People
- I. Hu
Organizations
- Stanford University