Quoting parameter-estimation results

Document #:

LIGO-T1500597-v1

Document type:

T - Technical notes

Abstract:

Measuring the properties of a gravitational-wave signal is a question of parameter estimation. The end result of these studies is a set of samples drawn from the posterior probability distribution. The posterior contains all the information that we have, but may not be easily digestible; in many cases, it is desirable to quote summary statistics to concisely describe findings. In most practical cases, it is impossible to compress a complete description of a distribution down to a single number; therefore, any point estimate could miss key pieces of information. Here, we discuss various possibilities for summary statistics. We also include a discussion of how to quote systematic errors on parameter estimates. While there is no perfect answer, our suggestion is to use $X^{+Y}_{-Z}$, where $X$ is the median $Y$ and $Z$ are estimates for the statistical error (measurement precision) from the bounds of a symmetric credible interval, and then add in estimates for systematic error from the range of $X$, $Y$ and $Z$, which could be presented as $X^{+Y\pm y}_{-Z\pm z}$ or $(X\pm x)^{+Y\pm y}_{-Z\pm z}$.

Files in Document:

• Manuscript (pointEstimate.pdf, 151.9 kB)

Topics:

• Data Analysis

Authors:

• Christopher Berry

Notes and Changes:

Files in CBC SVN under /pe/papers/pointEstimatesTechNote/