We develop a technique for constructing confidence lower bound for the parameter of the selected population in selection problems. This technique is then applied to selection problems relating to normal distribution or exponential distribution cases. Examples are provided to illustrate the performance of the proposed methods.