ANOVA for Some Non-Normal Data by Inverting Reliability Estimators

Standard analysis of variance assumes observations are normally distributed within groups. This paper develops some analysis of variance tests for data which are Bernoulli, Poisson, exponential, or geometric distributed within groups. The tests are shown in Table 1. For natural exponential family data with conjugate priors for the distribution of means, reliability estimators directly estimate the posterior shrinkage. Using the linear posterior expectation induced by conjugate prior, a method is developed to construct an analysis of variance test by determining an appropriate transformation of a reliability estimator. The sampling distribution of the transformed reliability estimator under the assumption of group mean equality is derived to construct an appropriate test statistic. This method is used to invert the generalized KR21 estimators of Foster (2021) for some non-normal data, and it is also shown that the standard analysis of variance F-test statistic can be transformed into a consistent reliability estimator under the same assumptions. A limited simulation study shows that the inverted KR21 test has, in some scenarios, higher power than a standard analysis of variance or a generalized linear model analysis of variance.

On the Notion of Reproducibility and Its Full Implementation to Natural Exponential Families

Let F=Fθ:θ∈Θ⊂R be a family of probability distributions indexed by a parameter θ and let X1,⋯,Xn be i.i.d. r.v.’s with L(X1)=Fθ∈F. Then, F is said to be reproducible if for all θ∈Θ and n∈N, there exists a sequence (αn)n≥1 and a mapping gn:Θ→Θ,θ⟼gn(θ) such that L(αn∑i=1nXi)=Fgn(θ)∈F. In this paper, we prove that a natural exponential family F is reproducible iff it possesses a variance function which is a power function of its mean. Such a result generalizes that of Bar-Lev and Enis (1986, The Annals of Statistics) who proved a similar but partial statement under the assumption that F is steep as and under rather restricted constraints on the forms of αn and gn(θ). We show that such restrictions are not required. In addition, we examine various aspects of reproducibility, both theoretically and practically, and discuss the relationship between reproducibility, convolution and infinite divisibility. We suggest new avenues for characterizing other classes of families of distributions with respect to their reproducibility and convolution properties .

On two extensions of the canonical Feller–Spitzer distribution

We introduce two extensions of the canonical Feller–Spitzer distribution from the class of Bessel densities, which comprise two distinct stochastically decreasing one-parameter families of positive absolutely continuous infinitely divisible distributions with monotone densities, whose upper tails exhibit a power decay. The densities of the members of the first class are expressed in terms of the modified Bessel function of the first kind, whereas the members of the second class have the densities of their Lévy measure given by virtue of the same function. The Laplace transforms for both these families possess closed–form representations in terms of specific hypergeometric functions. We obtain the explicit expressions by virtue of the particular parameter value for the moments of the distributions considered and establish the monotonicity of the mean, variance, skewness and excess kurtosis within the families. We derive numerous properties of members of these classes by employing both new and previously known properties of the special functions involved and determine the variance function for the natural exponential family generated by a member of the second class.

KR20 and KR21 for Some Nondichotomous Data (It’s Not Just Cronbach’s Alpha)

This article presents some equivalent forms of the common Kuder–Richardson Formula 21 and 20 estimators for nondichotomous data belonging to certain other exponential families, such as Poisson count data, exponential data, or geometric counts of trials until failure. Using the generalized framework of Foster (2020), an equation for the reliability for a subset of the natural exponential family have quadratic variance function is derived for known population parameters, and both formulas are shown to be different plug-in estimators of this quantity. The equivalent Kuder–Richardson Formulas 20 and 21 are given for six different natural exponential families, and these match earlier derivations in the case of binomial and Poisson data. Simulations show performance exceeding that of Cronbach’s alpha in terms of root mean square error when the formula matching the correct exponential family is used, and a discussion of Jensen’s inequality suggests explanations for peculiarities of the bias and standard error of the simulations across the different exponential families.

Location and scale behaviour of the quantiles of a natural exponential family

Let P0 be a probability on the real line generating a natural exponential family (Pt)t∈ℝ. Fix α in (0, 1). We show that the property that Pt((−∞, t)) ≤ α ≤ Pt((−∞, t]) for all t implies that there exists a number μα such that P0 is the Gaussian distribution N(μα, 1). In other terms, if for all t, the number t is a quantile of Pt associated to some threshold α ∈ (0, 1), then the exponential family must be Gaussian. The case α = 1∕2, i.e. when t is always a median of Pt, has been considered in Letac et al. [Statist. Prob. Lett. 133 (2018) 38–41]. Analogously let Q be a measure on [0, ∞) generating a natural exponential family (Q−t)t>0. We show that Q−t([0, t−1)) ≤ α ≤ Q−t([0, t−1]) for all t > 0 implies that there exists a number p = pα > 0 such that Q(dx) ∝ xp−1dx, and thus Q−t has to be a gamma law with parameters p and t.

Lévy processes time-changed by the first-exit time of the inverse Gaussian subordinator

This paper deals with a characterization of the first-exit time of the inverse Gaussian subordinator in terms of natural exponential family. This leads us to characterize, by means its variance function, the class of L?vy processes time-changed by the first-exit time of the inverse Gaussian subordinator.