Limit Theorems for Fisher-Score Change Processes
Abstract
Change analysis concerned with fluctuation of the data (in accordance with probability distributions fitted to a whole sample) from nonstationarity (changes in the parameters of probability distributions). To detect change over time in a sequence of observations one forms for various transformations of the data sample change processes on (0,1); the transformations are called data score functions . One can choose non-parametric score functions which detect changes of location, scale, skewness, etc. in the probability distribution of the observation When a parametric model is available for the distribution of each observation one can detect changes in the parameter values by transforming the data by parametric score functions which we call Fisher-score functions. This paper studies the asymptotic distributions (under the null hypothesis of no change) of Fisher-score change processes which are cusums of scored data. They are related to cuscore processes or cumulative score processes. Fisher-score change processes; Limit theorems.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1992
- Accession Number
- ADA257278
Entities
People
- Emanuel Parzen
- Lajos Horvath
Organizations
- Texas A&M University