-
Probability Based Independence Sampler for Bayesian Quantitative Learning in Graphical Log-Linear Marginal Models
Authors:
Ioannis Ntzoufras,
Claudia Tarantola,
Monia Lupparelli
Abstract:
Bayesian methods for graphical log-linear marginal models have not been developed in the same extent as traditional frequentist approaches. In this work, we introduce a novel Bayesian approach for quantitative learning for such models. These models belong to curved exponential families that are difficult to handle from a Bayesian perspective. Furthermore, the likelihood cannot be analytically expr…
▽ More
Bayesian methods for graphical log-linear marginal models have not been developed in the same extent as traditional frequentist approaches. In this work, we introduce a novel Bayesian approach for quantitative learning for such models. These models belong to curved exponential families that are difficult to handle from a Bayesian perspective. Furthermore, the likelihood cannot be analytically expressed as a function of the marginal log-linear interactions, but only in terms of cell counts or probabilities.
Posterior distributions cannot be directly obtained, and MCMC methods are needed. Finally, a well-defined model requires parameter values that lead to compatible marginal probabilities. Hence, any MCMC should account for this important restriction. We construct a fully automatic and efficient MCMC strategy for quantitative learning for graphical log-linear marginal models that handles these problems. While the prior is expressed in terms of the marginal log-linear interactions, we build an MCMC algorithm that employs a proposal on the probability parameter space. The corresponding proposal on the marginal log-linear interactions is obtained via parameter transformation.
By this strategy, we achieve to move within the desired target space. At each step, we directly work with well-defined probability distributions.
Moreover, we can exploit a conditional conjugate setup to build an efficient proposal on probability parameters. The proposed methodology is illustrated by a simulation study and a real dataset.
△ Less
Submitted 3 July, 2018;
originally announced July 2018.
-
Conditional and marginal relative risk parameters for a class of recursive regression graph models
Authors:
Monia Lupparelli
Abstract:
In linear regression modelling the distortion of effects after marginalizing over variables of the conditioning set has been widely studied in several contexts. For Gaussian variables, the relationship between marginal and partial regression coefficients is well-established and the issue is often addressed as a result of W. G. Cochran. Possible generalizations beyond the linear Gaussian case have…
▽ More
In linear regression modelling the distortion of effects after marginalizing over variables of the conditioning set has been widely studied in several contexts. For Gaussian variables, the relationship between marginal and partial regression coefficients is well-established and the issue is often addressed as a result of W. G. Cochran. Possible generalizations beyond the linear Gaussian case have been developed, nevertheless the case of discrete variables is still challenging, in particular in medical and social science settings. A multivariate regression framework is proposed for binary data with regression coefficients given by the logarithm of relative risks and a multivariate Relative Risk formula is derived to define the relationship between marginal and conditional relative risks. The method is illustrated through the analysis of the morphine data in order to assess the effect of preoperative oral morphine administration on the postoperative pain relief.
△ Less
Submitted 5 May, 2018;
originally announced May 2018.
-
Causal inference for binary non-independent outcomes
Authors:
Monia Lupparelli,
Alessandra Mattei
Abstract:
Causal inference on multiple non-independent outcomes raises serious challenges, because multivariate techniques that properly account for the outcome's dependence structure need to be considered. We focus on the case of binary outcomes framing our discussion in the potential outcome approach to causal inference. We define causal effects of treatment on joint outcomes introducing the notion of pro…
▽ More
Causal inference on multiple non-independent outcomes raises serious challenges, because multivariate techniques that properly account for the outcome's dependence structure need to be considered. We focus on the case of binary outcomes framing our discussion in the potential outcome approach to causal inference. We define causal effects of treatment on joint outcomes introducing the notion of product outcomes. We also discuss a decomposition of the causal effect on product outcomes into intrinsic and extrinsic causal effects, which respectively provide information on treatment effect on the intrinsic (product) structure of the product outcomes and on the outcomes' dependence structure. We propose a log-mean linear regression approach for modeling the distribution of the potential outcomes, which is particularly appealing because all the causal estimands of interest and the decomposition into intrinsic and extrinsic causal effects can be easily derived by model parameters. The method is illustrated in two randomized experiments concerning (i) the effect of the administration of oral pre-surgery morphine on pain intensity after surgery; and (ii) the effect of honey on nocturnal cough and sleep difficulty associated with childhood upper respiratory tract infections.
△ Less
Submitted 10 May, 2018; v1 submitted 19 October, 2017;
originally announced October 2017.
-
Log-mean linear regression models for binary responses with an application to multimorbidity
Authors:
Monia Lupparelli,
Alberto Roverato
Abstract:
In regression models for categorical data a linear model is typically related to the response variables via a transformation of probabilities called the link function. We introduce an approach based on two link functions for binary data named log-mean (LM) and log-mean linear (LML), respectively. The choice of the link function plays a key role for the interpretation of the model, and our approach…
▽ More
In regression models for categorical data a linear model is typically related to the response variables via a transformation of probabilities called the link function. We introduce an approach based on two link functions for binary data named log-mean (LM) and log-mean linear (LML), respectively. The choice of the link function plays a key role for the interpretation of the model, and our approach is especially appealing in terms of interpretation of the effects of covariates on the association of responses. Similarly to Poisson regression, the LM and LML regression coefficients of single outcomes are log-relative risks, and we show that the relative risk interpretation is maintained also in the regressions of the association of responses. Furthermore, certain collections of zero LML regression coefficients imply that the relative risks for joint responses factorize with respect to the corresponding relative risks for marginal responses. This work is motivated by the analysis of a dataset obtained from a case-control study aimed to investigate the effect of HIV-infection on multimorbidity, that is simultaneous presence of two or more noninfectious commorbidities in one patient.
△ Less
Submitted 16 May, 2016; v1 submitted 2 October, 2014;
originally announced October 2014.
-
Log-mean linear models for binary data
Authors:
Alberto Roverato,
Monia Lupparelli,
Luca La Rocca
Abstract:
This paper introduces a novel class of models for binary data, which we call log-mean linear models. The characterizing feature of these models is that they are specified by linear constraints on the log-mean linear parameter, defined as a log-linear expansion of the mean parameter of the multivariate Bernoulli distribution. We show that marginal independence relationships between variables can be…
▽ More
This paper introduces a novel class of models for binary data, which we call log-mean linear models. The characterizing feature of these models is that they are specified by linear constraints on the log-mean linear parameter, defined as a log-linear expansion of the mean parameter of the multivariate Bernoulli distribution. We show that marginal independence relationships between variables can be specified by setting certain log-mean linear interactions to zero and, more specifically, that graphical models of marginal independence are log-mean linear models. Our approach overcomes some drawbacks of the existing parameterizations of graphical models of marginal independence.
△ Less
Submitted 14 December, 2012; v1 submitted 28 September, 2011;
originally announced September 2011.
-
Latent Markov model for longitudinal binary data: An application to the performance evaluation of nursing homes
Authors:
Francesco Bartolucci,
Monia Lupparelli,
Giorgio E. Montanari
Abstract:
Performance evaluation of nursing homes is usually accomplished by the repeated administration of questionnaires aimed at measuring the health status of the patients during their period of residence in the nursing home. We illustrate how a latent Markov model with covariates may effectively be used for the analysis of data collected in this way. This model relies on a not directly observable Mar…
▽ More
Performance evaluation of nursing homes is usually accomplished by the repeated administration of questionnaires aimed at measuring the health status of the patients during their period of residence in the nursing home. We illustrate how a latent Markov model with covariates may effectively be used for the analysis of data collected in this way. This model relies on a not directly observable Markov process, whose states represent different levels of the health status. For the maximum likelihood estimation of the model we apply an EM algorithm implemented by means of certain recursions taken from the literature on hidden Markov chains. Of particular interest is the estimation of the effect of each nursing home on the probability of transition between the latent states. We show how the estimates of these effects may be used to construct a set of scores which allows us to rank these facilities in terms of their efficacy in taking care of the health conditions of their patients. The method is used within an application based on data concerning a set of nursing homes located in the Region of Umbria, Italy, which were followed for the period 2003--2005.
△ Less
Submitted 17 August, 2009;
originally announced August 2009.
-
Chain graph models of multivariate regression type for categorical data
Authors:
Giovanni M. Marchetti,
Monia Lupparelli
Abstract:
We discuss a class of chain graph models for categorical variables defined by what we call a multivariate regression chain graph Markov property. First, the set of local independencies of these models is shown to be Markov equivalent to those of a chain graph model recently defined in the literature. Next we provide a parametrization based on a sequence of generalized linear models with a multivar…
▽ More
We discuss a class of chain graph models for categorical variables defined by what we call a multivariate regression chain graph Markov property. First, the set of local independencies of these models is shown to be Markov equivalent to those of a chain graph model recently defined in the literature. Next we provide a parametrization based on a sequence of generalized linear models with a multivariate logistic link function that captures all independence constraints in any chain graph model of this kind.
△ Less
Submitted 13 July, 2011; v1 submitted 11 June, 2009;
originally announced June 2009.
-
Parameterizations and fitting of bi-directed graph models to categorical data
Authors:
Monia Lupparelli,
Giovanni M. Marchetti,
Wicher P. Bergsma
Abstract:
We discuss two parameterizations of models for marginal independencies for discrete distributions which are representable by bi-directed graph models, under the global Markov property. Such models are useful data analytic tools especially if used in combination with other graphical models. The first parameterization, in the saturated case, is also known as the multivariate logistic transformatio…
▽ More
We discuss two parameterizations of models for marginal independencies for discrete distributions which are representable by bi-directed graph models, under the global Markov property. Such models are useful data analytic tools especially if used in combination with other graphical models. The first parameterization, in the saturated case, is also known as the multivariate logistic transformation, the second is a variant that allows, in some (but not all) cases, variation independent parameters. An algorithm for maximum likelihood fitting is proposed, based on an extension of the Aitchison and Silvey method.
△ Less
Submitted 9 January, 2008;
originally announced January 2008.
