From O-C to SCUSUM+

Volume 22 number 2 (1993)

Download this article (pdf)

Grant Foster

AAVSO, Cambridge, MA

Abstract

Standard O-C (Observed-Computed) analysis assumes that the residual deviations of observed times from linearity are independent random variables. As Koen and Lombard (1993) point out, when a star's period is not constant but fluctuates about a constant mean value, this assumption leads to false confirmation that the period is changing. They supply an alternate test, CUSUM (CUmulative SUMs), which amounts to a series of scaled comparisons of "before" and "after" series averages. It can be extended by computing a separate scale factor for each individual CUSUM, yielding the SCUSUM test. This can be further extended to the SCUSUM+ test by including the effect of lag-1 correlation of the data.