The UU-test for statistical modeling of unimodal data

Deciding on the unimodality of a dataset is an important problem in data analysis and statistical modeling. It allows to obtain knowledge about structure dataset, ie. whether points have been generated by probability distribution with single or more than one peaks. Such very useful for several problems, such as deciding number clusters determining unimodal projections. We propose technique called UU-test (Unimodal Uniform test) decide one-dimensional dataset. The method operates empirical cumulative density function (ecdf) attempts build piecewise linear approximation ecdf that models sufficiently sense corresponding each segment follows uniform distribution. A unique feature this approach case unimodality, it also provides model form Mixture Model. present experimental results order assess ability perform comparisons well-known dip-test approach. In addition, datasets we evaluate Models provided proposed using test set log-likelihood two-sample Kolmogorov-Smirnov (KS) test.