Showing papers on "Resampling published in 1975"

Arguments for Fisher's Permutation Test

Abstract: The problem of statistical comparison of two distributions, continuous as well as discrete, is considered. Very slight and reasonable modifications of traditional parameteric models, e.g. `normal distributions with equal variances', are shown to result in permutation tests, only. Fisher's permutations test is shown to have optimum properties which mean a good merit for its practical use. Further, an accurate method of determining the $p$-value of Fisher's test is proposed.

A permutation test for periodicities in short, noisy time series

Abstract: We developed a distribution-free, nonparametric statistical test (the permutation-rank test) for verifying the exiscence of significant peaks in power spectral ordinates calculated on noisy, short, time series. In this paper, we demonstrate the application of the test in the case of periodic series generated by the in numero perturbation of the amplitude of a cosine wave of known, single, constant frequency. The test was able to detect the underlying oscillation when it was amplitude-corrupted by noise generated by a standard uniform distribution, with a noise-to-signal ratio of 1 or less.

Randomization and statistical inference

Abstract: An expository discussion is presented of the arguments concerning the role of randomization in statistical inference from the results of a randomized experiment. Then some comparisons of the power of the Fisher randomization test to the normal theory F-test for the case of an unpaired, two-group, randomized experiment are given..

Significance Testing on Randomized Mean Differences with a Small Computer

Abstract: Summary A method of data analysis is described in which the determination of exact probabilities associated with randomized mean differences between two independent samples may be programmed on a small computer. Application of the enumerative and probabilistic features of this method to specific randomization test analyses performed by other investigators revealed the presence of errors in several previously published reports.

Resampling of hypotheses after negative instances

Abstract: Ninety subjects were run under each of three conditions in a conjunctive concept identification experiment. For one problem, a subject was presented a repetition of the initial positive instance on the trial on which resampling would occur if the subject had not started with the correct hypothesis. For another problem, the subject was presented a new positive instance on that trial; and for a third problem, he was presented a negative instance on that trial. Hypothesis samples were assessed using blank trials following resampling. It was found that a repetition of the original positive instance and a negative instance resulted in about equal problem difficulty, while presentation of a new positive instance produced the best performance. Blank trial assessments corroborated the finding. It was concluded that memory for the initial focus was important in the resampling process.