James Collins

Publication Date


Advisor(s) - Committee Chair

Richard Hackney, Thomas Bohuski, Karen Hackney, N.F. Six

Degree Program

Department of Physics and Astronomy

Degree Type

Master of Science


Five astronomical data analysis methods were evaluated, using computer simulation techniques, for minimum detection levels, periodicity detection, and parameter estimation sensitivities. The evaluation of these methods was accomplished using three basic waveforms with the addition of ten “random” noise samples from a Gaussian distribution of known parameters. The chi-square and t-test methods were evaluated in detecting low-amplitude variability and brightness level shifts. Results indicate both methods have a strong dependence on the estimation of the standard deviation (σ) where a 0.1 change in σ will affect detection rates by ten to twenty-five percent at the ninety-five percent confidence level. The amplitude of variation must exceed four times σ of the sample before reliable detection will result. Two data smoothing techniques were investigated for estimating the rates of change of the variations. Tests indicate that the variation slope must equal or exceed six times σ of the sample with at least six points defining the slope in the case of the least squares method or four points while using the three point averaging as a preprocess to the least squares method before slope estimates fall within ± 10 percent of actual rates. Periodicity detection tests using autocorrelation show results of detection within ±1 point for the test samples where the σ of the random noise was equal to the basic waveform amplitude for uniformly spaced data which contains a minimum of one and one-half periods of variation.


Astrophysics and Astronomy | Physical Sciences and Mathematics | Physics