Testing that distributions are close
نویسندگان
چکیده
Given two distributions over an n element set, we wish to check whether these distributions are statistically close by only sampling. We give a sublinear algorithm which uses O(n2/3ǫ−4 logn) independent samples from each distribution, runs in time linear in the sample size, makes no assumptions about the structure of the distributions, and distinguishes the cases when the distance between the distributions is small (less than max( ǫ 2 32 3 √ n , ǫ 4 √ n )) or large (more than ǫ) in L1-distance. We also give an Ω(n2/3ǫ−2/3) lower bound. Our algorithm has applications to the problem of checking whether a given Markov process is rapidly mixing. We develop sublinear algorithms for this problem as well.
منابع مشابه
Comparing Mean Vectors Via Generalized Inference in Multivariate Log-Normal Distributions
Abstract In this paper, we consider the problem of means in several multivariate log-normal distributions and propose a useful method called as generalized variable method. Simulation studies show that suggested method has a appropriate size and power regardless sample size. To evaluation this method, we compare this method with traditional MANOVA such that the actual sizes of the two methods ...
متن کاملTracking Interval for Doubly Censored Data with Application of Plasma Droplet Spread Samples
Doubly censoring scheme, which includes left as well as right censored observations, is frequently observed in practical studies. In this paper we introduce a new interval say tracking interval for comparing the two rival models when the data are doubly censored. We obtain the asymptotic properties of maximum likelihood estimator under doubly censored data and drive a statistic for testing the ...
متن کاملBayesian Estimation of Reliability of the Electronic Components Using Censored Data from Weibull Distribution: Different Prior Distributions
The Weibull distribution has been widely used in survival and engineering reliability analysis. In life testing experiments is fairly common practice to terminate the experiment before all the items have failed, that means the data are censored. Thus, the main objective of this paper is to estimate the reliability function of the Weibull distribution with uncensored and censored data by using B...
متن کاملComparing the Shape Parameters of Two Weibull Distributions Using Records: A Generalized Inference
The Weibull distribution is a very applicable model for the lifetime data. For inference about two Weibull distributions using records, the shape parameters of the distributions are usually considered equal. However, there is not an appropriate method for comparing the shape parameters in the literature. Therefore, comparing the shape parameters of two Weibull distributions is very important. I...
متن کاملTesting a Point Null Hypothesis against One-Sided for Non Regular and Exponential Families: The Reconcilability Condition to P-values and Posterior Probability
In this paper, the reconcilability between the P-value and the posterior probability in testing a point null hypothesis against the one-sided hypothesis is considered. Two essential families, non regular and exponential family of distributions, are studied. It was shown in a non regular family of distributions; in some cases, it is possible to find a prior distribution function under which P-va...
متن کاملRange Distributions of Low-energy Nitrogen and Oxygen Ions in Silicon (RESEARCH NOTE)
The range distributions of low-energy nitrogen and oxygen (2-3 keV) ions is silicon are measured and compared with these available in theories. The nitrogen distribution is very close to a Gaussian distribution as predicted by theory. The oxygen profile however, indicates a surface localized peak along with a shoulder and a long tail into the sample. The surface peak is beleived to he the resul...
متن کامل