Note on distribution free testing for discrete distributions (1401.0609v1)

Abstract: The paper proposes one-to-one transformation of the vector of components ${Y_{in}}{i=1}^m$ of Pearson's chi-square statistic, [Y{in}=\frac{\nu_{in}-np_i}{\sqrt{np_i}},\qquad i=1,\ldots,m,] into another vector ${Z_{in}}{i=1}^m$, which, therefore, contains the same "statistical information," but is asymptotically distribution free. Hence any functional/test statistic based on ${Z{in}}{i=1}^m$ is also asymptotically distribution free. Natural examples of such test statistics are traditional goodness-of-fit statistics from partial sums $\sum{I\leq k}Z_{in}$. The supplement shows how the approach works in the problem of independent interest: the goodness-of-fit testing of power-law distribution with the Zipf law and the Karlin-Rouault law as particular alternatives.