Finite mixture regression model with random effects: application to neonatal hospital length of stay

A two-component mixture regression model that allows simultaneously for heterogeneity and dependency among observations is proposed. By specifying random effects explicitly in the linear predictor of the mixture probability and the mixture components, parameter estimation is achieved by maximising the corresponding best linear unbiased prediction type log-likelihood. Approximate residual maximum likelihood estimates are obtained via an EM algorithm in the manner of generalised linear mixed model (GLMM). The method can be extended to a g-component mixture regression model with the component density from the exponential family, leading to the development of the class of finite mixture GLMM. For illustration, the method is applied to analyse neonatal length of stay (LOS). It is shown that identification of pertinent factors that influence hospital LOS can provide important information for health care planning and resource allocation. (C) 2002 Elsevier Science B.V. All rights reserved.

Likelihood-based estimation of the regression model with scrambled responses

A significant problem in the collection of responses to potentially sensitive questions, such as relating to illegal, immoral or embarrassing activities, is non-sampling error due to refusal to respond or false responses. Eichhorn & Hayre (1983) suggested the use of scrambled responses to reduce this form of bias. This paper considers a linear regression model in which the dependent variable is unobserved but for which the sum or product with a scrambling random variable of known distribution, is known. The performance of two likelihood-based estimators is investigated, namely of a Bayesian estimator achieved through a Markov chain Monte Carlo (MCMC) sampling scheme, and a classical maximum-likelihood estimator. These two estimators and an estimator suggested by Singh, Joarder & King (1996) are compared. Monte Carlo results show that the Bayesian estimator outperforms the classical estimators in almost all cases, and the relative performance of the Bayesian estimator improves as the responses become more scrambled.

A modified discrete random pore model allowing for different initial surface reactivity

We investigate here a modification of the discrete random pore model [Bhatia SK, Vartak BJ, Carbon 1996;34:1383], by including an additional rate constant which takes into account the different reactivity of the initial pore surface having attached functional groups and hydrogens, relative to the subsequently exposed surface. It is observed that the relative initial reactivity has a significant effect on the conversion and structural evolution, underscoring the importance of initial surface chemistry. The model is tested against experimental data on chemically controlled char oxidation and steam gasification at various temperatures. It is seen that the variations of the reaction rate and surface area with conversion are better represented by the present approach than earlier random pore models. The results clearly indicate the improvement of model predictions in the low conversion region, where the effect of the initially attached functional groups and hydrogens is more significant, particularly for char oxidation. It is also seen that, for the data examined, the initial surface chemistry is less important for steam gasification as compared to the oxidation reaction. Further development of the approach must also incorporate the dynamics of surface complexation, which is not considered here.

A zero-augmented gamma mixed model for longitudinal data with many zeros

In many occupational safety interventions, the objective is to reduce the injury incidence as well as the mean claims cost once injury has occurred. The claims cost data within a period typically contain a large proportion of zero observations (no claim). The distribution thus comprises a point mass at 0 mixed with a non-degenerate parametric component. Essentially, the likelihood function can be factorized into two orthogonal components. These two components relate respectively to the effect of covariates on the incidence of claims and the magnitude of claims, given that claims are made. Furthermore, the longitudinal nature of the intervention inherently imposes some correlation among the observations. This paper introduces a zero-augmented gamma random effects model for analysing longitudinal data with many zeros. Adopting the generalized linear mixed model (GLMM) approach reduces the original problem to the fitting of two independent GLMMs. The method is applied to evaluate the effectiveness of a workplace risk assessment teams program, trialled within the cleaning services of a Western Australian public hospital.

A Mixture model with random-effects components for clustering correlated gene-expression profiles

Motivation: The clustering of gene profiles across some experimental conditions of interest contributes significantly to the elucidation of unknown gene function, the validation of gene discoveries and the interpretation of biological processes. However, this clustering problem is not straightforward as the profiles of the genes are not all independently distributed and the expression levels may have been obtained from an experimental design involving replicated arrays. Ignoring the dependence between the gene profiles and the structure of the replicated data can result in important sources of variability in the experiments being overlooked in the analysis, with the consequent possibility of misleading inferences being made. We propose a random-effects model that provides a unified approach to the clustering of genes with correlated expression levels measured in a wide variety of experimental situations. Our model is an extension of the normal mixture model to account for the correlations between the gene profiles and to enable covariate information to be incorporated into the clustering process. Hence the model is applicable to longitudinal studies with or without replication, for example, time-course experiments by using time as a covariate, and to cross-sectional experiments by using categorical covariates to represent the different experimental classes. Results: We show that our random-effects model can be fitted by maximum likelihood via the EM algorithm for which the E(expectation) and M(maximization) steps can be implemented in closed form. Hence our model can be fitted deterministically without the need for time-consuming Monte Carlo approximations. The effectiveness of our model-based procedure for the clustering of correlated gene profiles is demonstrated on three real datasets, representing typical microarray experimental designs, covering time-course, repeated-measurement and cross-sectional data. In these examples, relevant clusters of the genes are obtained, which are supported by existing gene-function annotation. A synthetic dataset is considered too.

Multi-level zero-inflated Poisson regression modelling of correlated count data with excess zeros

Count data with excess zeros relative to a Poisson distribution are common in many biomedical applications. A popular approach to the analysis of such data is to use a zero-inflated Poisson (ZIP) regression model. Often, because of the hierarchical Study design or the data collection procedure, zero-inflation and lack of independence may occur simultaneously, which tender the standard ZIP model inadequate. To account for the preponderance of zero counts and the inherent correlation of observations, a class of multi-level ZIP regression model with random effects is presented. Model fitting is facilitated using an expectation-maximization algorithm, whereas variance components are estimated via residual maximum likelihood estimating equations. A score test for zero-inflation is also presented. The multi-level ZIP model is then generalized to cope with a more complex correlation structure. Application to the analysis of correlated count data from a longitudinal infant feeding study illustrates the usefulness of the approach.

Stochastic structure and individual-tree growth models

The majority of past and current individual-tree growth modelling methodologies have failed to characterise and incorporate structured stochastic components. Rather, they have relied on deterministic predictions or have added an unstructured random component to predictions. In particular, spatial stochastic structure has been neglected, despite being present in most applications of individual-tree growth models. Spatial stochastic structure (also called spatial dependence or spatial autocorrelation) eventuates when spatial influences such as competition and micro-site effects are not fully captured in models. Temporal stochastic structure (also called temporal dependence or temporal autocorrelation) eventuates when a sequence of measurements is taken on an individual-tree over time, and variables explaining temporal variation in these measurements are not included in the model. Nested stochastic structure eventuates when measurements are combined across sampling units and differences among the sampling units are not fully captured in the model. This review examines spatial, temporal, and nested stochastic structure and instances where each has been characterised in the forest biometry and statistical literature. Methodologies for incorporating stochastic structure in growth model estimation and prediction are described. Benefits from incorporation of stochastic structure include valid statistical inference, improved estimation efficiency, and more realistic and theoretically sound predictions. It is proposed in this review that individual-tree modelling methodologies need to characterise and include structured stochasticity. Possibilities for future research are discussed. (C) 2001 Elsevier Science B.V. All rights reserved.

Modelling the effect of fibropapilloma disease on the somatic growth dynamics of Hawaiian green sea turtles

The effect of the tumour-forming disease, fibropapillomatosis, on the somatic growth dynamics of green turtles resident in the Pala'au foraging grounds (Moloka'i, Hawai'i) was evaluated using a Bayesian generalised additive mixed modelling approach. This regression model enabled us to account for fixed effects (fibropapilloma tumour severity), nonlinear covariate functional form (carapace size, sampling year) as well as random effects due to individual heterogeneity and correlation between repeated growth measurements on some turtles. Somatic growth rates were found to be nonlinear functions of carapace size and sampling year but were not a function of low-to-moderate tumour severity. On the other hand, growth rates were significantly lower for turtles with advanced fibropapillomatosis, which suggests a limited or threshold-specific disease effect. However, tumour severity was an increasing function of carapace size-larger turtles tended to have higher tumour severity scores, presumably due to longer exposure of larger (older) turtles to the factors that cause the disease. Hence turtles with advanced fibropapillomatosis tended to be the larger turtles, which confounds size and tumour severity in this study. But somatic growth rates for the Pala'au population have also declined since the mid-1980s (sampling year effect) while disease prevalence and severity increased from the mid-1980s before levelling off by the mid-1990s. It is unlikely that this decline was related to the increasing tumour severity because growth rates have also declined over the last 10-20 years for other green turtle populations resident in Hawaiian waters that have low or no disease prevalence. The declining somatic growth rate trends evident in the Hawaiian stock are more likely a density-dependent effect caused by a dramatic increase in abundance by this once-seriously-depleted stock since the mid-1980s. So despite increasing fibropapillomatosis risk over the last 20 years, only a limited effect on somatic growth dynamics was apparent and the Hawaiian green turtle stock continues to increase in abundance.

Testing the niche apportionment hypothesis with parasite communities: is random assortment always the rule?

Niche apportionment models have only been applied once to parasite communities. Only the random assortment model (RA), which indicates that species abundances are independent from each other and that interspecific competition is unimportant, provided a good fit to 3 out of 6 parasite communities investigated. The generality of this result needs to be validated, however. In this study we apply 5 niche apportionment models to the parasite communities of 14 fish species from the Great Barrier Reef. We determined which model fitted the data when using either numerical abundance or biomass as an estimate of parasite abundance, and whether the fit of niche apportionment models depends on how the parasite community is defined (e.g. ecto, endoparasites or all parasites considered together). The RA model provided a good fit for the whole community of parasites in 7 fish species when using biovolume (as a surrogate of biomass) as a measure of species abundance. The RA model also fitted observed data when ecto- and endoparasites were considered separately, using abundance or biovolume, but less frequently. Variation in fish sizes among species was not associated with the probability of a model fitting the data. Total numerical abundance and biovolume of parasites were not related across host species, suggesting that they capture different aspects of abundance. Biovolume is not only a better measurement to use with niche-orientated models, it should also be the preferred descriptor to analyse parasite community structure in other contexts. Most of the biological assumptions behind the RA model, i.e. randomness in apportioning niche space, lack of interspecific competition, independence of abundance among different species, and species with variable niches in changeable environments, are in accordance with some previous findings on parasite communities. Thus, parasite communities may generally be unsaturated with species, with empty niches, and interspecific interactions may generally be unimportant in determining parasite community structure.

Optimal crossover designs for logistic regression models in pharmacodynamics

Pharmacodynamics (PD) is the study of the biochemical and physiological effects of drugs. The construction of optimal designs for dose-ranging trials with multiple periods is considered in this paper, where the outcome of the trial (the effect of the drug) is considered to be a binary response: the success or failure of a drug to bring about a particular change in the subject after a given amount of time. The carryover effect of each dose from one period to the next is assumed to be proportional to the direct effect. It is shown for a logistic regression model that the efficiency of optimal parallel (single-period) or crossover (two-period) design is substantially greater than a balanced design. The optimal designs are also shown to be robust to misspecification of the value of the parameters. Finally, the parallel and crossover designs are combined to provide the experimenter with greater flexibility.

Review of recent results on excursion set models

Many images consist of two or more 'phases', where a phase is a collection of homogeneous zones. For example, the phases may represent the presence of different sulphides in an ore sample. Frequently, these phases exhibit very little structure, though all connected components of a given phase may be similar in some sense. As a consequence, random set models are commonly used to model such images. The Boolean model and models derived from the Boolean model are often chosen. An alternative approach to modelling such images is to use the excursion sets of random fields to model each phase. In this paper, the properties of excursion sets will be firstly discussed in terms of modelling binary images. Ways of extending these models to multi-phase images will then be explored. A desirable feature of any model is to be able to fit it to data reasonably well. Different methods for fitting random set models based on excursion sets will be presented and some of the difficulties with these methods will be discussed.

The kinetics of NO and N2O reduction over coal chars in fluidised-bed combustion

This paper presents a comprehensive and critical review of the mechanisms and kinetics of NO and N2O reduction reaction with coal chars under fluidised-bed combustion conditions (FBC). The heterogeneous reactions of NO and N2O with char/carbon surface have been well recognised as the most important processes in reducing both NOx and N2O in situ FBC. Compared to NO-carbon reactions in FBC, the reactions of N2O with chars have been relatively less understood and studied. Beginning with the overall reaction schemes for both NO and N2O reduction, the paper extensively discusses the reaction mechanisms including the effects of active surface sites. Generally, NO- and N2O-carbon reactions follow a series of step reactions. However, questions remain concerning the role of adsorbed phases of NO and N2O, and the behaviour of different surface sites. Important kinetics factors such as the rate expressions, kinetics parameters as well as the effects of surface area and pore structure are discussed in detail. The main factors influencing the reduction of NO and N2O in FBC conditions are the chemical and physical properties of chars, and the operating parameters of FBC such as temperature, presence of CO, O-2 and pressure. It is shown that under similar conditions, N2O is more readily reduced on the char surface than NO. Temperature was found to be a very important parameter in both NO and N2O reduction. It is generally agreed that both NO- and N2O-carbon reactions follow first-order reaction kinetics with respect to the NO and N2O concentrations. The kinetic parameters for NO and N2O reduction largely depend on the pore structure of chars. The correlation between the char surface area and the reactivities of NO/N2O-char reactions is considered to be of great importance to the determination of the reaction kinetics. The rate of NO reduction by chars is strongly enhanced by the presence of CO and O-2, but these species may not have significant effects on the rate of N2O reduction. However, the presence of these gases in FBC presents difficulties in the study of kinetics since CO cannot be easily eliminated from the carbon surface. In N2O reduction reactions, ash in chars is found to have significant catalytic effects, which must be accounted for in the kinetic models and data evaluation. (C) 1997 Elsevier Science Ltd.

Reactivity of chars and carbons: New insights through molecular modeling

A new model proposed for the gasification of chars and carbons incorporates features of the turbostratic nanoscale structure that exists in such materials. The model also considers the effect of initial surface chemistry and different reactivities perpendicular to the edges and to the faces of the underlying crystallite planes comprising the turbostratic structure. It may be more realistic than earlier models based on pore or grain structure idealizations when the carbon contains large amounts of crystallite matter. Shrinkage of the carbon particles in the chemically controlled regime is also possible due to the random complete gasification of crystallitic planes. This mechanism can explain observations in the literature of particle size reduction. Based on the model predictions, both initial surface chemistry and the number of stacked planes in the crystallites strongly influence the reactivity and particle shrinkage. Its test results agree well with literature data on the air-oxidation of Spherocarb and show that it accurately predicts the variation of particle size with conversion. Model parameters are determined entirely from rate measurements.

Who fails to complete tuberculosis treatment? Temporal trends and risk factors for treatment interruption in a community-based directly observed therapy programme in a rural district of South Africa

SETTING: Hlabisa Tuberculosis Programme, Hlabisa, South Africa. OBJECTIVE: To determine trends in and risk factors for interruption of tuberculosis treatment. METHODS: Data were extracted from the control programme database starting in 1991. Temporal trends in treatment interruption are described; independent risk factors for treatment interruption were determined with a multiple logistic regression model, and Kaplan-Meier survival curves for treatment interruption were constructed for patients treated in 1994-1995. RESULTS: Overall 629 of 3610 surviving patients (17%) failed to complete treatment; this proportion increased from 11% (n = 79) in 1991/1992 to 22% (n = 201) in 1996. Independent risk factors for treatment interruption were diagnosis between 1994-1996 compared with 1991-1393 (odds ratio [OR] 1.9, 95% confidence interval [CT] 1.6-2.4); human immunodeficiency virus (HIV) positivity compared with HIV negativity (OR 1.8, 95% CI 1.4-2.4); supervised by village clinic compared with community health worker (OR 1.9, 95% CI 1.4-2.6); and male versus female sex (OR 1.3, 95% CI 1.1-1.6). Few patients interrupted treatment during the first 2 weeks, and the treatment interruption rate thereafter was constant at 1% per 14 days. CONCLUSIONS: Frequency of treatment interruption from this programme has increased recently. The strongest risk factor was year of diagnosis, perhaps reflecting the impact of an increased caseload on programme performance. Ensuring adherence to therapy in communities with a high level of migration remains a challenge even within community-based directly observed therapy programmes.

Variation in mortality rates in Australia: correlation with Indigenous status, remoteness and socio-economic deprivation

Background The aim of this study was to study ecological correlations between age-adjusted all-cause mortality rates in Australian statistical divisions and (1) the proportion of residents that self-identify as Indigenous, (2) remoteness, and (3) socio-economic deprivation. Methods All-cause mortality rates for 57 statistical divisions were calculated and directly standardized to the 1997 Australian population in 5-year age groups using Australian Bureau of Statistics (ABS) data. The proportion of residents who self-identified as Indigenous was obtained from the 1996 Census. Remoteness was measured using ARIA (Accessibility and Remoteness Index for Australia) values. Socioeconomic deprivation was measured using SEIFA (Socio-Economic index for Australia) values from the ABS. Results Age-standardized all-cause mortality varies twofold from 5.7 to 11.3 per 1000 across Australian statistical divisions. Strongest correlation was between Indigenous status and mortality (r = 0.69, p < 0.001). correlation between remoteness and mortality was modest (r = 0.39, p = 0.002) as was correlation between socio-economic deprivation and mortality (r = -0.42, p = 0.001). Excluding the three divisions with the highest mortality, a multiple regression model using the logarithm of the adjusted mortality rate as the dependent variable showed that the partial correlation (and hence proportion of the variance explained) for Indigenous status was 0.03 (9 per cent; p = 0.03), for SEIFA score was -0.17 (3 per cent; p = 0.22); and for remoteness was -0.22 (5 per cent; p = 0.13). Collectively, the three variables studied explain 13 per cent of the variability in mortality. Conclusions Ecological correlation exists between all-cause mortality, Indigenous status, remoteness and disadvantage across Australia. The strongest correlation is with indigenous status, and correlation with all three characteristics is weak when the three statistical divisions with the highest mortality rates are excluded. intervention targeted at these three statistical divisions could reduce much of the variability in mortality in Australia.