On the I(q) Test Statistic for Spatial Dependence: Finite Sample Standardization and Properties
نویسندگان
چکیده
One of the most widely used tests for spatial dependence is Moran’s (1950) I test. The power of the test will depend on the extent to which the spatial-weights matrix employed in computing the Moran I test statistic properly specifies existing interaction links between spatial units. Empirical researchers are often unsure about the use of a particular spatial-weights matrix. In light of this Prucha (2011) introduced the I(q) test statistic. This test statistic combines quadratic forms based on several, say q, spatial-weights matrices, while at the same time allows for a proper controlling of the size of the test. In this paper, we first introduce a finite-sample standardized version of the I(q) test. We then perform a Monte Carlo study to explore the finite-sample performance of the I(q) tests. For comparison, the Monte Carlo study also reports on the finite-sample performance of Moran I tests as well as on Moran I tests performed in sequence. Des statistiques du test I(q) pour la dépendance spatiale: harmonisation des échantillons finis et propriétés RÉSUMÉ un des tests les plus répandus de la dépendance spatiale est celui de Moran. La puissance de ce test est tributaire de la mesure dans laquelle la matrice de pondération spatiale, employée pour calculer correctement les statistiques du test, spécifie correctement les liens d’interaction existants entre unités spatiales. Les chercheurs empiriques éprouvent souvent des incertitudes en ce qui concerne l’emploi d’une certaine matrice de pondération spatiale. Pour cette raison, Prucha (2011) a introduit la statistique de test I(q), assurant la combinaison de formes quadratiques sur plusieurs matrices de pondération spatiale, par exemple q. Dans la présente communication, nous introduisons une version harmonisée aux éléments finis de ce test, et nous présentons un compte rendu sur une étude Monte Carlo
منابع مشابه
Bootstrapping the Box-pierce Q Test: a Robust Test of Uncorrelatedness
This paper describes a test of the null hypothesis that the first K autocorrelations of a covariance stationary time series are zero in the presence of statistical dependence. The test is based on the BoxPierce Q statistic with bootstrap-based P-values. The bootstrap is implemented using a double blocks-of-blocks procedure with prewhitening. The finite sample performance of the bootstrap Q test...
متن کاملPRELIMINARY AND INCOMPLETE DRAFT Bootstrapping the Box - Pierce Q test : A Robust Test of Uncorrelatedness
This paper considers a test of the null hypothesis that the first K autocorrelations of a covariance stationary time series are zero in the presence of statistical dependence. The test is based on the BoxPierce Q statistic with bootstrap-based critical values. The bootstrap is implemented using a double blocks of blocks procedure with prewhitening. The finite sample performance of the bootstrap...
متن کاملDo Stock Returns Follow a Finite Variance Distribution?
In this paper we propose a test statistic to discriminate between models with finite variance and models with infinite variance. The test statistic is the ratio of the sample standard deviation and the sample interquartile range. Both asymptotic and finite sample properties of the test statistic are discussed. We show that the test has good power properties against infinite-variance distributio...
متن کاملWhich is the Preferred Measure of Heterogeneity
Heterogeneity assessment is unavoidable under meta-analysis as it guides the path to choose appropriate synthesizing method. Its consideration further strengthens the drawn inferences. In the literature, there are various measures of heterogeneity, such as, Q statistic, H2 statistic, HM2 statistic, R2 statistic and I2 statistic. The present work mainly aimed to appraise all existing methods and...
متن کاملOptimal kernel choice for large-scale two-sample tests
Given samples from distributions p and q, a two-sample test determines whether to reject the null hypothesis that p = q, based on the value of a test statistic measuring the distance between the samples. One choice of test statistic is the maximum mean discrepancy (MMD), which is a distance between embeddings of the probability distributions in a reproducing kernel Hilbert space. The kernel use...
متن کامل