Resumo

COMAS ARIAS, Niuman; CATALA GONZALEZ, Belarmino  e  ORO DOSOUTO, Oscar. Goodness of fit test for distance distribution in categorical data sequences. Rev cuba cienc informat [online]. 2021, vol.15, n.2, pp. 62-76.  Epub 01-Jun-2021. ISSN 2227-1899.

Randomness analysis in categorical sequences is relevant for the study of Markov processes, system realibity, big data, data encryption and evaluation of pseudo-random number generators. Various approaches exist in order to appraise the randomness phenomena, they lead to a variety of tests such as the “Diehard” test battery, the test U01 and the NIST Statistical Test Suite. The behavior of categorical sequences was studied and understood as a discrete time chronological series. It was proved that the geometric distribution is the expected distribution (theoretical distribution) for distances between successes random sequences. The observed distance distribution was compared to the theoretical distribution by goodness of fit test based on chi-square statistic. The test algorithm was implemented as javascript module for web statistical packages checking its sensibility to various no random behavior including the periodical character of successes, blocking, autocorrelation and Markov processes existence. Test convergence and robustness were studied by means of simulation in computer, discovering little deviations in proportion of the significant cases that indicate the existence of inherent biased in chi-square test.

Palavras-chave : Categorical sequences; randomness; goodness of fit test.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )