Search

The Finite Selection Model (FSM) was developed by Carl Morris in the 1970s for the design of the RAND Health Insurance Experiment (HIE) (Morris 1979, Newhouse et al. 1993), one of the largest and most comprehensive social science experiments conducted in the U.S. The idea behind the FSM is that each treatment group takes its turns selecting units in a fair and random order to optimize a common assignment criterion. At each of its turns, a treatment group selects the available unit that maximally improves the combined quality of its resulting group of units in terms of the criterion. In the HIE and beyond, we revisit, formalize, and extend the FSM as a general tool for experimental design. Leveraging the idea of D-optimality, we propose and analyze a new selection criterion in the FSM. The FSM using the D-optimal selection function has no tuning parameters, is affine invariant, and when appropriate, retrieves several classical designs such as randomized block and matched-pair designs. For multi-arm experiments, we propose algorithms to generate a fair and random selection order of treatments. We demonstrate FSM's performance in a case study based on the HIE and in ten randomized studies from the health and social sciences. On average, the FSM achieves 68% better covariate balance than complete randomization and 56% better covariate balance than rerandomization in a typical study. We recommend the FSM be considered in experimental design for its conceptual simplicity, efficiency, and robustness.

The original Finite Selection Model (FSM) was developed in the 1970s to enhance the design of the RAND Health Insurance Experiment (HIE; Newhouse et al. 1993). At the time of its development by Carl Morris (Morris 1979), there were fundamental computational limitations to make the method widely available for practitioners. Today, as randomized experiments increasingly become more common, there is a need for implementing experimental designs that are randomized, balanced, robust, and easily applicable to several treatment groups. To help address this problem, we revisit the original FSM under the potential outcome framework for causal inference and provide its first readily available software implementation. In this paper, we provide an introduction to the FSM and a step-by-step guide for its use in R.

Rgbp is an R package that provides estimates and verifiable confidence intervals for random effects in two-level conjugate hierarchical models for overdispersed Gaussian, Poisson, and Binomial data. Rgbp models aggregate data from k independent groups summarized by observed sufficient statistics for each random effect, such as sample means, possibly with covariates. Rgbp uses approximate Bayesian machinery with unique improper priors for the hyper-parameters, which leads to good repeated sampling coverage properties for random effects. A special feature of Rgbp is an option that generates synthetic data sets to check whether the interval estimates for random effects actually meet the nominal confidence levels. Additionally, Rgbp provides inference statistics for the hyper-parameters, e.g., regression coefficients.

A Beta-Binomial-Logit model is a Beta-Binomial model with covariate information incorporated via a logistic regression. Posterior propriety of a Bayesian Beta-Binomial-Logit model can be data-dependent for improper hyper-prior distributions. Various researchers in the literature have unknowingly used improper posterior distributions or have given incorrect statements about posterior propriety because checking posterior propriety can be challenging due to the complicated functional form of a Beta-Binomial-Logit model. We derive data-dependent necessary and sufficient conditions for posterior propriety within a class of hyper-prior distributions that encompass those used in previous studies.

This review traces the evolution of theory that started when Charles Stein in 1955 [In Proc. 3rd Berkeley Sympos. Math. Statist. Probab. I (1956) 197--206, Univ. California Press] showed that using each separate sample mean from $k\ge3$ Normal populations to estimate its own population mean $\mu_i$ can be improved upon uniformly for every possible $\mu=(\mu_1,...,\mu_k)'$. The dominating estimators, referred to here as being "Model-I minimax," can be found by shrinking the sample means toward any constant vector. Admissible minimax shrinkage estimators were derived by Stein and others as posterior means based on a random effects model, "Model-II" here, wherein the $\mu_i$ values have their own distributions. Section 2 centers on Figure 2, which organizes a wide class of priors on the unknown Level-II hyperparameters that have been proved to yield admissible Model-I minimax shrinkage estimators in the "equal variance case." Putting a flat prior on the Level-II variance is unique in this class for its scale-invariance and for its conjugacy, and it induces Stein's harmonic prior (SHP) on $\mu_i$., Comment: Published in at http://dx.doi.org/10.1214/11-STS363 the Statistical Science (http://www.imstat.org/sts/) by the Institute of Mathematical Statistics (http://www.imstat.org)

Rejoinder of "Estimating Random Effects via Adjustment for Density Maximization" by C. Morris and R. Tang [arXiv:1108.3234], Comment: Published in at http://dx.doi.org/10.1214/11-STS349REJ the Statistical Science (http://www.imstat.org/sts/) by the Institute of Mathematical Statistics (http://www.imstat.org)

We develop and evaluate point and interval estimates for the random effects $\theta_i$, having made observations $y_i|\theta_i\stackrel{\m athit{ind}}{\sim}N[\theta_i,V_i],i=1,...,k$ that follow a two-level Normal hierarchical model. Fitting this model requires assessing the Level-2 variance $A\equiv\operatorname {Var}(\theta_i)$ to estimate shrinkages $B_i\equiv V_i/(V_i+A)$ toward a (possibly estimated) subspace, with $B_i$ as the target because the conditional means and variances of $\theta_i$ depend linearly on $B_i$, not on $A$. Adjustment for density maximization, ADM, can do the fitting for any smooth prior on $A$. Like the MLE, ADM bases inferences on two derivatives, but ADM can approximate with any Pearson family, with Beta distributions being appropriate because shrinkage factors satisfy $0\le B_i\le1$. Our emphasis is on frequency properties, which leads to adopting a uniform prior on $A\ge0$, which then puts Stein's harmonic prior (SHP) on the $k$ random effects. It is known for the "equal variances case" $V_1=...=V_k$ that formal Bayes procedures for this prior produce admissible minimax estimates of the random effects, and that the posterior variances are large enough to provide confidence intervals that meet their nominal coverages. Similar results are seen to hold for our approximating "ADM-SHP" procedure for equal variances and also for the unequal variances situations checked here. For shrinkage coefficient estimation, the ADM-SHP procedure allows an alternative frequency interpretation. Writing $L(A)$ as the likelihood of $B_i$ with $i$ fixed, ADM-SHP estimates $B_i$ as $\hat{B_i}=V_i/(V_i+\hat{A})$ with $\hat{A}\equiv \operatorname {argmax}(A*L(A))$. This justifies the term "adjustment for likelihood maximization," ALM., Comment: Published in at http://dx.doi.org/10.1214/10-STS349 the Statistical Science (http://www.imstat.org/sts/) by the Institute of Mathematical Statistics (http://www.imstat.org)

Brad Efron's paper [arXiv:0808.0572] has inspired a return to the ideas behind Bayes, frequency and empirical Bayes. The latter preferably would not be limited to exchangeable models for the data and hyperparameters. Parallels are revealed between microarray analyses and profiling of hospitals, with advances suggesting more decision modeling for gene identification also. Then good multilevel and empirical Bayes models for random effects should be sought when regression toward the mean is anticipated., Comment: Published in at http://dx.doi.org/10.1214/08-STS236D the Statistical Science (http://www.imstat.org/sts/) by the Institute of Mathematical Statistics (http://www.imstat.org)