Length tests for goodness of fit
基于曲线长度的拟合优度检验
The length of the graph of an estimate of the distribution function is used as a measure of goodness of fit. The asymptotic theory of tests which are based on such a measure is studied and, in addition, tables of critical values for these tests are given. In the context of time series analysis, the proposed test rejects the null hypothesis of white noise whenever the length of the linearly interpolated cumulative periodogram, or of some smoothed version of it, is too large.