Comparison of Linear Trends in Time Series Data using Regression Analysis.
Abstract
Under the assumption that each of two sets of time series data contains a linear trend and stationary Gaussian autocorrelated noise, equations are developed to test the null hypothesis that the trends are the same. Due to the frequency of occurrence of red noise (the noise elements form a linear first-order Markov chain) in geophysical data, this type of noise is treated in detail. The consequences of assuming the wrong red noise model for the data are investigated. The techniques developed in the paper are then applied to atmospheric density data derived from rocket measurements, and the effects of different red noise models are shown. As a result the density noise is found to be red noise with a lag one-kilometer autocorrelation coefficient of about 0.8 and a standard deviation of approximately 1.085% of the true density. (author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1971
- Accession Number
- AD0734331
Entities
People
- Bruce. T. Miers
- Elton P. Avara
Organizations
- Atmospheric Sciences Laboratory