Abstract
The Newcomb-Benford probability distribution is becoming very popular in many areas using statistics, notably in fraud detection. In such contexts, it is important to be able to determine if a data set arises from this distribution while controlling the risk of a Type 1 error, i.e. falsely identifying a fraud, and a Type 2 error, i.e. not detecting that a fraud occurred. The statistical tool to do this work is a goodness-of-fit test. For the Newcomb-Benford distribution, the most popular such test is Pearson's chi-square test whose power, related to the Type 2 error, is known to be weak. Consequently, other tests have been recently introduced. The goal of the present work is to build new goodness-of-fit tests for this distribution, based on the smooth test principle. These tests are then compared to some of their competitors. It turns out that the proposals of the paper are globally preferable to existing tests and should be seriously considered in fraud detection contexts, among others