Monotone Likelihood Research Articles

An inequality for log-concave functions and its use in the study of failure rates

Abstract We establish here an integral inequality for real log-concave functions, which can be viewed as an average monotone likelihood property. This inequality is then applied to examine the monotonicity of failure rates.

Simultaneous directional inference

Abstract We consider the problem of inference on the signs of n>1 parameters. We aim to provide 1−α post hoc confidence bounds on the number of positive and negative (or non-positive) parameters, with a simultaneous guarantee, for all subsets of parameters. We suggest to start by using the data to select the direction of the hypothesis test for each parameter; then, adjust the p-values of the one-sided hypotheses for the selection, and use the adjusted p-values for simultaneous inference on the selected n one-sided hypotheses. The adjustment is straightforward assuming the p-values of one-sided hypotheses have densities with monotone likelihood ratio, and are mutually independent. We show the bounds we provide are tighter (often by a great margin) than existing alternatives, and that they can be obtained by at most a polynomial time. We demonstrate their usefulness in the evaluation of treatment effects across studies or subgroups. Specifically, we provide a tight lower bound on the number of studies which are beneficial, as well as on the number of studies which are harmful (or non-beneficial), and in addition conclude on the effect direction of individual studies, while guaranteeing that the probability of at least one wrong inference is at most 0.05.

Stochastic ordering results on the duration of the gambler’s ruin game

AbstractIn the classical gambler’s ruin problem, the gambler plays an adversary with initial capitals z and $a-z$ , respectively, where $a>0$ and $0< z < a$ are integers. At each round, the gambler wins or loses a dollar with probabilities p and $1-p$ . The game continues until one of the two players is ruined. For even a and $0<z\leq {a}/{2}$ , the family of distributions of the duration (total number of rounds) of the game indexed by $p \in [0,{\frac{1}{2}}]$ is shown to have monotone (increasing) likelihood ratio, while for ${a}/{2} \leq z<a$ , the family of distributions of the duration indexed by $p \in [{\frac{1}{2}}, 1]$ has monotone (decreasing) likelihood ratio. In particular, for $z={a}/{2}$ , in terms of the likelihood ratio order, the distribution of the duration is maximized over $p \in [0,1]$ by $p={\frac{1}{2}}$ . The case of odd a is also considered in terms of the usual stochastic order. Furthermore, as a limit, the first exit time of Brownian motion is briefly discussed.

Comparing information in general monotone decision problems

I study the value of information in monotone decision problems with potentially multidimensional action spaces. As a criterion for comparing information structures, I develop a condition called monotone quasi-garbling, which involves adding reversely monotone noise to an existing information structure. Specifically, this noise is more likely to return a higher signal in a lower state and a lower signal in a higher state. I show that monotone quasi-garbling is a necessary and sufficient condition for decision makers to obtain a higher ex-ante expected payoff. This new criterion refines the garbling condition by Blackwell (1951, 1953) and is equivalent to the accuracy condition by Lehmann (1988) under the monotone likelihood ratio property. To illustrate, I apply the result to problems in nonlinear monopoly pricing and optimal insurance.

Bayesian solution to the monotone likelihood in the standard mixture cure model

An advantage of the standard mixture cure model over an usual survival model is how it accounts for the population heterogeneity. It allows a joint estimation for the distribution related to the susceptible and non‐susceptible subjects. The estimation algorithm may provide coefficients when the likelihood cannot be maximized. This phenomenon is known as Monotone Likelihood (ML), common in survival and logistic regressions. The ML tends to appear in situations with small sample size, many censored times, many binary or unbalanced covariates. Particularly, it occurs when all uncensored cases correspond to one level of a binary covariate. The existing frequentist solution is an adaptation of the Firth correction, originally proposed to reduce bias of maximum likelihood estimates. It prevents estimates by penalizing the likelihood, with the penalty interpreted as the Bayesian Jeffreys prior. In this paper, the penalized likelihood of the standard mixture cure model is considered with different penalties (Bayesian priors). A Monte Carlo simulation study indicates good inference results, especially for balanced data sets. Finally, a real application involving a melanoma data illustrates the approach.

Risk Classification in Insurance Markets with Risk and Preference Heterogeneity

Abstract This paper studies a competitive model of insurance markets in which consumers are privately informed about their risk and risk preferences. We provide a characterization of the equilibria, which depend non-trivially on consumers’ type distribution, a desirable feature for policy analysis. The use of consumer characteristics for risk classification is modeled as the disclosure of a public informative signal. A novel property of signals, monotonicity, is shown to be necessary and sufficient for their release to be welfare improving for almost all consumer types. We also study the effect of changes to the risk distribution in the population as the result of demographic changes or policy interventions. We show that an increase in the risk distribution, according to the monotone likelihood ratio ordering of distribution, leads to lower utility for almost all consumer types. In contrast, the effect is ambiguous when considering the first-order stochastic dominance ordering.

Variable selection, monotone likelihood ratio and group sparsity

In the pivotal variable selection problem, we derive the exact nonasymptotic minimax selector over the class of all s-sparse vectors, which is also the Bayes selector with respect to the uniform prior. While this optimal selector is, in general, not realizable in polynomial time, we show that its tractable counterpart (the scan selector) attains the minimax expected Hamming risk to within factor 2, and is also exact minimax with respect to the probability of wrong recovery. As a consequence, we establish explicit lower bounds under the monotone likelihood ratio property and we obtain a tight characterization of the minimax risk in terms of the best separable selector risk. We apply these general results to derive necessary and sufficient conditions of exact and almost full recovery in the location model with light tail distributions and in the problem of group variable selection under Gaussian noise and under more general anisotropic sub-Gaussian noise. Numerical results illustrate our theoretical findings.

Sequential Detection of an Arbitrary Transient Change Profile by the FMA Test

This article addresses the sequential detection of transient changes by using the finite moving average (FMA) test. It is assumed that a change occurs at an unknown (but nonrandom) change point and the duration of postchange period is finite and known. We relax the assumption that the profile of a transient change is chosen so that the log-likelihood ratios of the observations are associated random variables (r.v.s) in the prechange mode. Hence, the profile of the transient change is arbitrary and it is not necessarily of constant sign for a distribution with monotone likelihood ratio. A new upper bound for the worst-case probability of false alarm is proposed. It is shown that the optimization of the window-limited cumulative sum (CUSUM) test again leads to the FMA test. Three particular transient changes are considered: in the Gaussian mean, in the Gaussian variance, and in the parameter of exponential distribution. In the first case, a comparison between the bounds for the FMA test operating characteristics and the exact operating characteristics calculated by numerical integration is used to estimate the sharpness of the bounds. In the second and third cases, special attention is paid to the calculation of the FMA distribution in the case of arbitrary profile. The method of convolution is used to solve the problem.

A Statistical Test for the Detection of Item Compromise Combining Responses and Response Times

AbstractA test of item compromise is presented which combines the test takers' responses and response times (RTs) into a statistic defined as the number of correct responses on the item for test takers with RTs flagged as suspicious. The test has null and alternative distributions belonging to the well‐known family of compound binomial distributions, is simple to calculate, and has results that are easy to interpret. It also demonstrated nearly perfect power for the detection of compromise with no more than 10 test takers with preknowledge of the more difficult and discriminating items in a set of empirical examples. For the easier and less discriminating items, the presence of some 20 test takers with preknowledge still sufficed. A test based on the reverse statistic of the total time by test takers with responses flagged as suspicious may seem a natural alternative but misses the property of a monotone likelihood ratio necessary to decide between a test that should be left or right sided.

Pseudo-Marginal Inference for CTMCs on Infinite Spaces via Monotonic Likelihood Approximations

Bayesian inference for Continuous-Time Markov chains (CTMCs) on countably infinite spaces is notoriously difficult because evaluating the likelihood exactly is intractable. One way to address this challenge is to first build a nonnegative and unbiased estimate of the likelihood—involving the matrix exponential of finite truncations of the true rate matrix—and then to use the estimates in a pseudo-marginal inference method. In this work, we show that we can dramatically increase the efficiency of this approach by avoiding the computation of exact matrix exponentials. In particular, we develop a general methodology for constructing an unbiased, nonnegative estimate of the likelihood using doubly-monotone matrix exponential approximations. We further develop a novel approximation in this family—the skeletoid—as well as theory regarding its approximation error and how that relates to the variance of the estimates used in pseudo-marginal inference. Experimental results show that our approach yields more efficient posterior inference for a wide variety of CTMCs. Supplementary materials for this article are available online.

On monotone likelihood ratio of stationary probabilities in bonus-malus systems

Abstract Bonus-malus system is an often used risk management tool in the insurance industry, and it is usually modeled with Markov chains. Under mild conditions it can be stated that the bonus-malus system converges to a unique stationary distribution in the long run. The maximum likelihood ratio property is a well-known statistical concept and we define it for the stationary distribution of a bonus-malus system. For two special cases we could justify it algebraically. For other cases we describe a numerical method with which we can test this property in any case. With the help of the described method, we checked this property for cases that appear in actuarial practice.

Dealing with the Problem of Monotone Likelihood in the Inflation of Estimated Effects in Clinical Studies. Comment on Hasegawa et al. Impact of Blood Type O on Mortality of Sepsis Patients: A Multicenter Retrospective Observational Study. Diagnostics 2020, 10, 826.

I read the press article from the Fujita Health University School of Medicine titled "Impact of Blood Type O on Mortality of Sepsis Patients: A Multicenter Retrospective Observational Study" [...].

On estimation for accelerated failure time models with small or rare event survival data

BackgroundSeparation or monotone likelihood may exist in fitting process of the accelerated failure time (AFT) model using maximum likelihood approach when sample size is small and/or rate of censoring is high (rare event) or there is at least one strong covariate in the model, resulting in infinite estimates of at least one regression coefficient.MethodsThis paper investigated the properties of the maximum likelihood estimator (MLE) of the regression parameters of the AFT models for small sample and/or rare-event situation and addressed the problems by introducing a penalized likelihood approach. The penalized likelihood function and the corresponding score equation is derived by adding a penalty term to the existing likelihood function, which was originally proposed by Firth (Biometrika, 1993) for the exponential family models. Further, a post-hoc adjustment of intercept and scale parameters is discussed keeping them out of penalization to ensure accurate prediction of survival probability. The penalized method was illustrated for the widely used log-location-scale family models such as Weibull, Log-normal and Log-logistic distributions and compared the models and methods uisng an extensive simulation study.ResultsThe simulation study, performed separately for each of the log-location-scale models, showed that Firth’s penalized likelihood succeeded to solve the problem of separation and achieve convergence, providing finite estimates of the regression coefficients, which are not often possible by the MLE. Furthermore, the proposed penalized method showed substantial improvement over MLE by providing smaller amount of bias, mean squared error (MSE), narrower confidence interval and reasonably accurate prediction of survival probabilities. The methods are illustrated using prostate cancer data with existence of separation, and results supported the simulation findings.ConclusionWhen sample size is small (≤ 50) or event is rare (i.e., censoring proportion is high) and/or there is any evidence of separation in the data, we recommend to use Firth’s penalized likelihood method for fitting AFT model.

A penalized likelihood approach for dealing with separation in count data regression model

Separation or monotone likelihood can be observed in fitting process of both Poisson or generalized Poisson (GP) regression, particularly in case of small and/or sparse count data, using maximum likelihood estimation (MLE) when one or more regression coefficients diverge to infinity. The study investigates the consequence of separation in the MLE based standard Poisson or GP model and addressed the problems by introducing a penalized likelihood approach. The penalized likelihood function is derived by adding a penalty term to the standard maximum likelihood function, which was originally proposed by Firth (1993) for reducing first order bias in MLE. The corresponding penalized likelihood score equation has shown to achieve convergence and provide finite estimate of the regression coefficient, which was not possible for the maximum likelihood method score equation. The simulation study, with different forms of separation, showed that penalized Poisson model or penalized GP outperform the standard Poisson and even the Zero-inflated Poisson (ZIP) in the presence of complete or quasi-complete separation by achieving convergence and providing finite estimate of the regression coefficients. Even in the presence of near-to-quasi-complete separation, which is very common in practice, the penalized method showed better results than the standard Poisson, GP and ZIP in all simulation scenarios. The method was illustrated using antenatal care data extracted from database of the Bangladesh demographic health survey 2018.

Simple contracts with adverse selection and moral hazard

We study a principal–agent model with moral hazard and adverse selection. Risk‐neutral agents with limited liability have arbitrary private information about the distribution of outputs and the cost of effort. We show that under a multiplicative separability condition, the optimal mechanism offers a single contract. This condition holds, for example, when output is binary. If the principal's payoff must also satisfy free disposal and the distribution of outputs has the monotone likelihood ratio property, the mechanism offers a single debt contract. Our results generalize if the output distribution is “close” to multiplicatively separable. Our model suggests that offering a single contract may be optimal in environments with adverse selection and moral hazard when agents are risk‐neutral and have limited liability.

Fair Cake Division Under Monotone Likelihood Ratios

This work develops algorithmic results for the classic cake-cutting problem in which a divisible, heterogeneous resource (modeled as a cake) needs to be partitioned among agents with distinct preferences. We focus on a standard formulation of cake cutting wherein each agent must receive a contiguous piece of the cake. Although multiple hardness results exist in this setup for finding fair/efficient cake divisions, we show that, if the value densities of the agents satisfy the monotone likelihood ratio property (MLRP), then strong algorithmic results hold for various notions of fairness and economic efficiency. Addressing cake-cutting instances with MLRP, first we develop an algorithm that finds cake divisions (with connected pieces) that are envy free, up to an arbitrary precision. The time complexity of our algorithm is polynomial in the number of agents and the bit complexity of an underlying Lipschitz constant. We obtain similar positive results for maximizing social, egalitarian, and Nash social welfare. Many distribution families bear MLRP. In particular, this property holds if all the value densities belong to any one of the following families: Gaussian (with the same variance), linear, Poisson, and exponential distributions, linear translations of any log-concave function. Hence, through MLRP, the current work obtains novel cake-cutting algorithms for multiple distribution families.

Modified score function for monotone likelihood in the semiparametric mixture cure model.

The cure fraction models are intended to analyze lifetime data from populations where some individuals are immune to the event under study, and allow a joint estimation of the distribution related to the cured and susceptible subjects, as opposed to the usual approach ignoring the cure rate. In situations involving small sample sizes with many censored times, the detection of nonfinite coefficients may arise via maximum likelihood. This phenomenon is commonly known as monotone likelihood (ML), occurring in the Cox and logistic regression models when many categorical and unbalanced covariates are present. An existing solution to prevent the issue is based on the Firth correction, originally developed to reduce the estimation bias. The method ensures finite estimates by penalizing the likelihood function. In the context of mixture cure models, the ML issue is rarely discussed in the literature; therefore, this topic can be seen as the first contribution of our paper. The second major contribution, not well addressed elsewhere, is the study of the ML issue in cure mixture modeling under the flexibility of a semiparametric framework to handle the baseline hazard. We derive the modified score function based on the Firth approach and explore finite sample size properties of the estimators via a Monte Carlo scheme. The simulation results indicate that the performance of coefficients related to the binary covariates are strongly affected to the imbalance degree. A real illustration, in the melanoma dataset, is discussed using a relatively novel data set collected in a Brazilian universityhospital.

Rational inattention and the monotone likelihood ratio property

This paper explores the connection between rational inattention and the monotone likelihood ratio property (MLRP). In many economic problems, agents prefer to take a higher action when there is a higher underlying state. It is often assumed that these agents have noisy signals that are strongly correlated with the underlying state, i.e. ordered by the MLRP. Separately, a large literature has recently explored the behavior of rationally inattentive agents. Under some common separability and smoothness assumptions on the class of attention costs, I show that agents will always acquire signals that are ordered by the MLRP within such economic problems if and only if they have attention costs proportional to entropy reduction. With any other attention costs, agents err regarding the relative benefit of actions at different states. Extending to games, I provide conditions under which players first acquire MLRP-ordered signals, and then choose higher actions given higher signals. I then apply these results to independent private-value auctions.

A New Three-Parameter Weibull Inverse Rayleigh Distribution: Theoretical Development and Applications

In this work, a three-parameter Weibull Inverse Rayleigh (WIR) distribution is proposed. The new WIR distribution is an extension of a one-parameter Inverse Rayleigh distribution that incorporated a transformation of the Weibull distribution and Log-logistic as quantile function. The statistical properties such as quantile function, order statistic, monotone likelihood ratio property, hazard, reverse hazard functions, moments, skewness, kurtosis, and linear representation of the new proposed distribution were studied theoretically. The maximum likelihood estimators cannot be derived in an explicit form. So we employed the iterative procedure called Newton Raphson method to obtain the maximum likelihood estimators. The Bayes estimators for the scale and shape parameters for the WIR distribution under squared error, Linex, and Entropy loss functions are provided. The Bayes estimators cannot be obtained explicitly. Hence we adopted a numerical approximation method known as Lindley's approximation in other to obtain the Bayes estimators. Simulation procedures were adopted to see the effectiveness of different estimators. The applications of the new WIR distribution were demonstrated on three real-life data sets. Further results showed that the new WIR distribution performed credibly well when compared with five of the related existing skewed distributions. It was observed that the Bayesian estimates derived performs better than the classical method.

Online Appendix for Security Design for Private Acquisitions

Online Appendix A to Security Design for Asset Acquisitions shows that seller debt securities can be optimal security designs when informed sellers, rather than uninformed buyers, make acquisition offers, and when acquirers are cash constrained. In contrast, acquirer overconfidence and seller private information are essential components of our analysis of seller debt financing. Online Appendix B to Security Design for Asset Acquisition' shows that the ratio condition, Assumption 4, is satisfied by all standard textbook distributions (supported by the non-negative real line) whenever the standard monotone likelihood ratio conditions (Assumptions 1 and 2) of security design are satisfied. The distributions families considered are Chi-Squared, Exponential, Gamma, Gompertz, Half Normal, Lognormal, and Weibull. General sufficient conditions for the satisfaction of Assumption 5 are also developed. The paper is available at https://ssrn.com/abstract=3731086.