Statistical Issues in the MACHO Project

The MACHO project is an ongoing survey project which collects photometric timeseries data on roughly 20 million stars in the LMC, SMC, and Galactic bulge. The data is irregularly time sampled, largely due to weather interruptions, and has noise which is non-stationary and non-Gaussian. The time series data is analyzed for microlensing events, rare brightenings of a star that result from an otherwise undetected massive object that passes close to the line of sight, thereby forming a gravitational lens. In addition to microlensing events, there are a much larger number of brightenings that result from intrinsic stellar variability. This background is only partially understood, since some classes of variable stars have received little study, and there are doubtless some classes yet to be discovered. Since the background can not be reliably simulated, the experiment must aim at a false alarm rate near zero, at the cost of reduced detection efficiency. At the same time, to achieve the scientific results at which it aims, the detection efficiency must be measured reliably.