A statistical approach to station discontinuance

The sequential probability ratio test (SPRT) is developed and applied to the station discontinuance problem. The decision whether to discontinue a station is dependent upon statistical considerations that include the error probabilities of accepting a model when it is incorrect as well as rejecting it when it is correct. Both error probabilities can be explicitly set by using the SPRT approach so that the decision rules are: accept model 1, accept model 2, continue to sample (cannot accept either). The error probability levels are a decision variable that depends upon the use of the data and the cost of collection. Procedures are developed for determining the expected benefits from sampling one additional time within the SPRT framework. The SPRT approach was applied to the Susquehanna River at Harrisburg, Pennsylvania.