Event studies have found several applications in financial and accounting researches as the standard method to assess the average effect of some types of announcement on stock prices. In a large sample of announcements, event window length can be standardized (fixed) across observations, because the errors from having too long or too short event window should have small impact in average by the Law of Large Numbers. But in small samples (in emerging markets) we cannot use this procedure. Here, we examine various potential rules for determining the length of an event window when looking at limited number of observations. We consider the announcements of adjustment of predicted earnings (by companies) in Tehran Stock Exchange during 1386 to 1388. To determining the length of time period affecting the market, tree methods are used. We also examine the relationship between the length of event window and the amount of unexpected earnings.