THE SIGN TEST ON AUTOREGRESSIVE DATA.
Abstract
The sign test is analyzed and shown to lose its distribution-free properties for dependent data of the form Xsubt = pXsubt-1 + Wsubt, t = 1, ..., N, O p < 1; (Wsubt) is a strictly stationary sequence of independent, identically distributed random variables with zero mean. Formulas for the mean and variance of the sign test statistic are derived and examples given for Xsubt being Gaussian and double-exponential. The Blum-Rosenblatt Central Limit Theorem for strong mixing processes is used to prove asymptotic normality of the sign test statistic in these two cases. Anomalous results of a Monte Carlo experiment are reported. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1965
- Accession Number
- AD0625272
Entities
People
- H. Rubin
- J. Gastwirth
- S. S. Wolff
Organizations
- Johns Hopkins University