Burst Gravitational Wave Data Analysis Methods:<br> Design, Development and Comparison

Open Access
Stuver, Amber Lynnell
Graduate Program:
Doctor of Philosophy
Document Type:
Date of Defense:
June 30, 2006
Committee Members:
  • Lee S Finn, Committee Chair
  • Pablo Laguna, Committee Member
  • Stephane Coutu, Committee Member
  • Steinn Sigurdsson, Committee Member
  • General Relativity
  • LIGO
  • simulation
  • signal processing
  • data analysis
  • gravitational waves
<p>The advent of interferometric gravitational wave observatories, such as LIGO (Laser Interferometric Gravitational-wave Observatory), have promise of opening a new spectrum in astronomy. Throughout history, each time humans have looked at the universe in a different spectrum they have made discoveries that were not anticipated. With this in mind, the search for short duration gravitational wave bursts from unanticipated sources is underway. Several data analysis methods have been developed to search for these bursts which will be buried in a sea of instrumental noise. However, investigations have not been undertaken to determine if these methods are fundamentally equivalent and, if not, what are their relative strengths and weaknesses.</p> <p>This work documents the results of asking the following questions of the burst gravitational wave data analysis methods: are the methods equivalent, and if they are not, what time-domain signal properties does each method favor? and are the time-domain characteristics of a signal more important to its detection by a data analysis method or are the frequency-domain characteristics more important?</p> <p>To determine if the burst data analysis methods are equivalent, they were applied to LIGO science data and the results from each method were ranked by strength as measured by each method. If the methods were equivalent, then each method would see the same events and rank them in approximately the same way. This was not the case. With this knowledge, simulations with adjustable parameters were injected into a well controlled background of white noise. The efficacy of each data analysis method was probed with respect to simulation signal type, frequency, duration and amplitude. The efficiency measurements from these simulations illuminated the signal preferences of each method and served as a baseline to investigate the effect of varying the signals' time-domain properties while holding their frequency-domain magnitude constant. This evaluated the relative importance of the time-domain versus frequency-domain properties for a signal's detection efficiency.</p> Ultimately, this work shows that while not fundamentally equivalent, these data analysis methods are complementary to each other in their relative strengths and features.