Asymptotic power of the two-sample test of Wilcoxon for Logistic shift alternatives, and comparison with simulation results

In this paper the power of the two-sample test of W ilcoxon for Logistic shift alternatives is computed by using two approximations, based on the asymptotic Normality of the test statistic. The sample sizes considered are m = n = 6, m = n= 10 and m = n = 15 . These approximations are compared with the results of an experimental determination by Monte Carlo techniques.     

Testing for the bivariate Poisson distribution

This paper studies goodness-of-fit tests for the bivariate Poisson distribution. Specifically, we propose and study several Cramér–von Mises type tests based on the empirical probability generating function. They are consistent against fixed alternatives for adequate choices of the weight function involved in their definition. They are also able to detect local alternatives converging to the null at a certain rate. The bootstrap can be used to consistently estimate the null distribution of the test statistics. A simulation study investigates the goodness of the bootstrap approximation and compares their powers for finite sample sizes. Extensions for testing goodness-of-fit for the multivariate Poisson distribution are also discussed.     

Nearly exact sample size calculation for powerful non‐randomized tests for differences between binomial proportions

In the case of two independent samples, it turns out that among the procedures taken in consideration, BOSCHLOO'S technique of raising the nominal level in the standard conditional test as far as admissible performs best in terms of power against almost all alternatives. The computational burden entailed in exact sample size calculation is comparatively modest for both the uniformly most powerful unbiased randomized and the conservative non‐randomized version of the exact Fisher‐type test. Computing these values yields a pair of bounds enclosing the exact sample size required for the Boschloo test, and it seems reasonable to replace the exact value with the middle of the corresponding interval. Comparisons between these mid‐N estimates and the fully exact sample sizes lead to the conclusion that the extra computational effort required for obtaining the latter is mostly dispensable. This holds also true in the case of paired binary data (McNemar setting). In the latter, the level‐corrected score test turns out to be almost as powerful as the randomized uniformly most powerful unbiased test and should be preferred to the McNemar–Boschloo test. The mid‐N rule provides a fairly tight upper bound to the exact sample size for the score test for paired proportions.     

On the Kuiper test for normality with mean and variance unknown

Summary A modified form of the Kuiper statistic V _n is developed for testing the composite hypothesis that a sample of size n comes from a normal population with unspecified mean and variance. Its distribution is derived using Monte Carlo methods. Power comparison with the adjusted Kuiper test proposed by L outer and K oerts [6] indicates that our test is superior with respect to certain alternatives.     

On the Power of GLS-Type Unit Root Tests

Elliott, Rothenberg and Stock (1996), (ERS), present a 'GLS' variant of the Dickey-Fuller (DF) unit root test. Their statistic is approximately point-optimal invariant at a chosen local alternative, and usually displays better finite sample power than the DF test. Following the usual efficiency motive for GLS estimation, the higher finite sample power of the ERS test has often been attributed to the greater accuracy of the estimate of the series' non-stochastic component under stationary alternatives close to the null. This paper shows that the GLS estimates of the non-stochastic component are not, in general, more accurate. The power gain arises from the fact that the GLS statistic's null distribution has a greater positive shift relative to the DF test, than its distribution under relevant alternatives, and this persists even when the GLS estimates of the non stochastics have higher variance than the OLS estimates.     

A SIMPLE TEST FOR UNIFORMITY

Abstract A new and very simple test for uniformity is proposed. An exact formula for the distribution under H₀ of the corresponding test statistic is derived. This formula is only suitable for computer-oriented use. For other circumstances a table of critical values is given. The power of the test is compared with that of two well-known alternatives: the χ²-test and the Kolmogorov-Smirnov test.     

Goodness-of-fit tests in linear EV regression with replications

This paper proposes a goodness-of-fit test for checking the adequacy of parametric forms of the regression error density functions in linear errors-in-variables regression models. Instead of assuming the distribution of the measurement error to be known, we assume that replications of the surrogates of the latent variables are available. The test statistic is based upon a weighted integrated squared distance between a nonparametric estimator and a semi-parametric estimator of the density functions of certain residuals. Under the null hypothesis, the test statistic is shown to be asymptotically normal. Consistency and local power results of the proposed test under fixed alternatives and local alternatives are also established. Finite sample performance of the proposed test is evaluated via simulation studies. A real data example is also included to demonstrate an application of the proposed test.     

Bayesian Calibration of p‐Values from Fisher's Exact Test

p‐Values are commonly transformed to lower bounds on Bayes factors, so‐called minimum Bayes factors. For the linear model, a sample‐size adjusted minimum Bayes factor over the class of g‐priors on the regression coefficients has recently been proposed (Held & Ott, The American Statistician 70(4), 335–341, 2016). Here, we extend this methodology to a logistic regression to obtain a sample‐size adjusted minimum Bayes factor for 2 × 2 contingency tables. We then study the relationship between this minimum Bayes factor and two‐sided p‐values from Fisher's exact test, as well as less conservative alternatives, with a novel parametric regression approach. It turns out that for all p‐values considered, the maximal evidence against the point null hypothesis is inversely related to the sample size. The same qualitative relationship is observed for minimum Bayes factors over the more general class of symmetric prior distributions. For the p‐values from Fisher's exact test, the minimum Bayes factors do on average not tend to the large‐sample bound as the sample size becomes large, but for the less conservative alternatives, the large‐sample behaviour is as expected.     

Goodness-of-fit tests for long memory moving average marginal density

This paper addresses the problem of fitting a known density to the marginal error density of a stationary long memory moving average process when its mean is known and unknown. In the case of unknown mean, when mean is estimated by the sample mean, the first order difference between the residual empirical and null distribution functions is known to be asymptotically degenerate at zero, and hence can not be used to fit a distribution up to an unknown mean. In this paper we show that by using a suitable class of estimators of the mean, this first order degeneracy does not occur. We also investigate the large sample behavior of tests based on an integrated square difference between kernel type error density estimators and the expected value of the error density estimator based on errors. The asymptotic null distributions of suitably standardized test statistics are shown to be chi-square with one degree of freedom in both cases of the known and unknown mean. In addition, we discuss the consistency and asymptotic power against local alternatives of the density estimator based test in the case of known mean. A finite sample simulation study of the test based on residual empirical process is also included.     

A weighted Kuiper statistic for goodness of fit

A variance-weighted Kuiper statistic for goodness of fit is studied. The exact finite sample distribution can be obtained through modification of Noé's (1972) algorithm. Asymptotic distribution theory for the statistic is available from Jaeschke (1979) and Eicker (1979), but this theory does not lead to useful approximations with finite sample sizes less than 100. Monte Carlo power studies demonstrate that the weighted Kuiper statistic is especially sensitive to alternatives that are not stochastically ordered relative to the postulated null distribution.     

On Tests for a Normal Mean with Known Coefficient of Variation

Hinkley (1977) derived two tests for testing the mean of a normal distribution with known coefficient of variation (c.v.) for right alternatives. They are the locally most powerful (LMP) and the conditional tests based on the ancillary statistic for μ. In this paper, the likelihood ratio (LR) and Wald tests are derived for the one‐ and two‐sided alternatives, as well as the two‐sided version of the LMP test. The performances of these tests are compared with those of the classical t, sign and Wilcoxon signed rank tests. The latter three tests do not use the information on c.v. Normal approximation is used to approximate the null distribution of the test statistics except for the t test. Simulation results indicate that all the tests maintain the type‐I error rates, that is, the attained level is close to the nominal level of significance of the tests. The power functions of the tests are estimated through simulation. The power comparison indicates that for one‐sided alternatives the LMP test is the best test whereas for the two‐sided alternatives the LR or the Wald test is the best test. The t, sign and Wilcoxon signed rank tests have lower power than the LMP, LR and Wald tests at various alternative values of μ. The power difference is quite large in several simulation configurations. Further, it is observed that the t, sign and Wilcoxon signed rank tests have considerably lower power even for the alternatives which are far away from the null hypothesis when the c.v. is large. To study the sensitivity of the tests for the violation of the normality assumption, the type I error rates are estimated on the observations of lognormal, gamma and uniform distributions. The newly derived tests maintain the type I error rates for moderate values of c.v.     

Testing parametric conditional distributions using the nonparametric smoothing method

This paper proposes a new goodness-of-fit test for parametric conditional probability distributions using the nonparametric smoothing methodology. An asymptotic normal distribution is established for the test statistic under the null hypothesis of correct specification of the parametric distribution. The test is shown to have power against local alternatives converging to the null at certain rates. The test can be applied to testing for possible misspecifications in a wide variety of parametric models. A bootstrap procedure is provided for obtaining more accurate critical values for the test. Monte Carlo simulations show that the test has good power against some common alternatives.     

Testing equality of variances in the analysis of repeated measurements

The problem of comparing the precisions of two instruments using repeated measurements can be cast as an extension of the Pitman-Morgan problem of testing equality of variances of a bivariate normal distribution. Hawkins (1981) decomposes the hypothesis of equal variances in this model into two subhypotheses for which simple tests exist. For the overall hypothesis he proposes to combine the tests of the subhypotheses using Fisher's method and empirically compares the component tests and their combination with the likelihood ratio test. In this paper an attempt is made to resolve some discrepancies and puzzling conclusions in Hawkins's study and to propose simple modifications.
The new tests are compared to the tests discussed by Hawkins and to each other both in terms of the finite sample power (estimated by Monte Carlo simulation) and theoretically in terms of asymptotic relative efficiencies.     

On distribution-free goodness-of-fit testing of exponentiality

There is a need to test the hypothesis of exponentiality against a wide variety of alternative hypotheses, across many areas of economics and finance. Local or contiguous alternatives are the closest alternatives against which it is still possible to have some power. Hence goodness-of-fit tests should have some power against all, or a huge majority, of local alternatives. Such tests are often based on nonlinear statistics, with a complicated asymptotic null distribution. Thus a second desirable property of a goodness-of-fit test is that its statistic will be asymptotically distribution free. We suggest a whole class of goodness-of-fit tests with both of these properties, by constructing a new version of empirical process that weakly converges to a standard Brownian motion under the hypothesis of exponentiality. All statistics based on this process will asymptotically behave as statistics from a standard Brownian motion and so will be asymptotically distribution free. We show the form of transformation is especially simple in the case of exponentiality. Surprisingly there are only two asymptotically distribution free versions of empirical process for this problem, and only this one has a convenient limit distribution. Many tests of exponentiality have been suggested based on asymptotically linear functionals from the empirical process. We illustrate none of these can be used as goodness-of-fit tests, contrary to some previous recommendations. Of considerable interest is that a selection of well-known statistics all lead to the same test asymptotically, with negligible asymptotic power against a great majority of local alternatives. Finally, we present an extension of our approach that solves the problem of multiple testing, both for exponentiality and for other, more general hypotheses.     

Experimental determination of the power functions of the two-sample rank tests of WILCOXON, VAN DER WAERDEN and TERRY by Monte Carlo techniques

In this paper the authors present the results of a sampling experiment to determine the power functions of the two-sample rank tests of WILCOXON, VAN DER WAERDEN and TERRY against shift alternatives for normal parent distributions, based on 2000 trials for each alternative. The sample sizes considered are m = n = 6 and m = n = 10. The powers of the three rank tests are compared with the power of the STUDENT t-test and with each other. The results indicate that in small samples (i) the power of the WILCOXON test is not much smaller than the power of the t-test and (ii) the normal scores tests are only slightly superior to the WILCOXON test, if at all.     

On Testing for the Nullity of Some Skewness Coefficients

Three tests for the skewness of an unknown distribution are derived for iid data. They are based on suitable normalization of estimators of some usual skewness coefficients. Their asymptotic null distributions are derived. The tests are next shown to be consistent and their power under some sequences of local alternatives is investigated. Their finite sample properties are also studied through a simulation experiment, and compared to those of the √ b ₂-test.     

A test for normality

Summary An analogue of the C ramer - von M ism W ²-statistic is given for testing the composite hypothesis of normality with unspecified parameters. Some Monte Carlo percentiles of this statistic are provided. A power comparison with the test developed in [4] shows that the present test is better against certain alternatives.     

Confidence bounds for the sensitivity lack of a less specific diagnostic test, without gold standard

We consider the problem of comparing two diagnostic tests based on a sample of paired test results without true state determinations, in cases where the second test can reasonably be assumed to be at least as specific as the first. For such cases, we provide two informative confidence bounds: A lower one for the prevalence times the sensitivity gain of the second test with respect to the first, and an upper one for the sensitivity of the first test. Neither conditional independence of the two tests nor perfectness of any of them needs to be assumed. An application of the proposed confidence bounds to a sample of 256 pairs of laboratory test results for toxigenic Clostridium difficile provides evidence for a dramatic sensitivity gain through first appropriately culturing C. difficile from stool samples before applying an enzyme-immuno-assay.