Author

Anderson

Date of Award

4-1982

Degree Name

Master of Arts

Department

Psychology

First Advisor

Dr. Bradley E. Huitema

Second Advisor

Dr. Dale Brethower

Third Advisor

Dr. R. Wayne Fuqua

Access Setting

Masters Thesis-Open Access

Abstract

Many methodologists have asserted that serial dependency is a common data characteristic of research employing single subject responses measured over time. This dependency, when present, invalidates usage of most conventional data analysis techniques and may necessitate use of Auto-Regressive Integrated Moving Average (ARIMA) models which accommodate serial dependency characteristics of those data. Results indicate that minimal levels of serial dependency are present in most areas, yet is higher in areas which sample behaviors at very long or short time intervals and the dependent data usually show linear trend or cycle. A review shows that once trends and cycles are accounted for, serial dependency is very uncommon. The results indicate that serial dependency is not as prevalent as many researchers have assumed it to be and the need for widespread use of ARIMA models in operant research is limited.

Included in

Psychology Commons

Share

COinS