Calculation of the effect estimate (in this case odds ratio) necessarily implies pooling the data together and producing an average estimate of risk. However, there might be situations when crude estimates are misleading. In particular, if there is a strong confounding in the data. There are several ways to deal with confounding in the data. At the study analysis stage, we can stratify the data into several groups based on the level of the confounding variable, calculate estimates of effect in each strata and then combine them by using a Mantel-Haenszel procedure. This statistic combines information across partial tables and enables you to calculate one common OR, as opposed to many for each strata.