This paper describes an upper tail point estimation method using the incomplete data by the maximum likelihood estimation method. In some data cases, it happensto occur that the estimates for upper tail points such as 100-year return periodcan be smaller than the observed maximum annual flood datum, even if the number of observed data is smaller than 100; this result of analysis cannot be acceptable from the practioners' viewpoint. Because, adopting such estimates in hydrologic frequency analysis may cause dangerous designs and plans for flood prevention. By weighting much to data in upper tail probability portion, such a phenomenonmay be reduced. Peaks over threshold method (POT) and probability weighted moment method (PWM) are included in such weighting methods. This paper deals with a weighting method by using incomplete data. An example for St. Mary's River data shows the superior points in such a treatment.