On a class of stopping times for M-estimators (Q797943): Difference between revisions

Let \(\theta_ n\) be an M-estimator for an unknown parameter \(\theta\) when the score function is \(\psi =\psi (x,\theta)\). Under some conditions it is known that \(\sqrt{n}(\theta_ n-\theta)\) is asymptotically normal. The author considers the problem of confidence sequences for \(\theta\) and studies the associated stopping time \(\tau_ c(\theta)=\inf\{n\geq m:\quad\sqrt{n}|\theta_ n-\theta| >c\},\) and discusses some limit results such as the law of the iterated logarithm, etc.