Of course, in MLE, if we collect more and more observations data, MLE will perform better and better. But is there a way to find a bound on parameter estimate errors in order to decide when to stop collect data/observations, say 1000 observations are great, but 500 observations is good enough... Thanks!