Bartlett type identities for martingales (Q1327830)

The Bartlett identities for moments and cumulants of log likelihood derivatives are a very powerful tool in likelihood inference, leading to some quite general results in that area. This paper shows that these identities also apply to martingales. As applications, the author gives a cumulant-based proof of the martingale central limit theorem and an algorithm for calculating approximate cumulants of least squares estimators in AR(1) processes.