144 research outputs found
A Common Platform for Graphical Models in R: The gRbase Package
No full textThe gRbase package is intended to set the framework for computer packages for data analysis using graphical models. The gRbase package is developed for the open source language, R, and is available for several platforms. The package is intended to be widely extendible and flexible so that package developers may implement further types of graphical models using the available methods. The gRbase package consists of a set of S version 3 classes and associated methods for representing data and models. The package is linked to the dynamicGraph package (Badsberg 2005), an interactive graphical user interface for manipulating graphs. In this paper, we show how these building blocks can be combined and integrated with inference engines in the special cases of hierarchical loglinear models. We also illustrate how to extend the package to deal with other types of graphical models, in this case the graphical Gaussian models.
deal: A Package for Learning Bayesian Networks
No full textdeal is a software package for use with R. It includes several methods for analysing data using Bayesian networks with variables of discrete and/or continuous types but restricted to conditionally Gaussian networks. Construction of priors for network parameters is supported and their parameters can be learned from data using conjugate updating. The network score is used as a metric to learn the structure of the network and forms the basis of a heuristic search strategy. deal has an interface to Hugin.
Gradvist ændrende sæsonvariation af kardiovaskulære sygdomme: - Et dansk nationalt kohorte studie
No full textThe influence of fatty fish on changes in waist circumference : modelled by mixed effect models and meta analysis
No full textGorst-Rasmussen et al. Respond to "Dietary Pattern Analysis"
No full textWe thank Imamura and Jacques (1) for their insightful commentary on our article (2), in which they go beyond the treelet transform (TT) to critically discuss the relevance of sparsity in dietary pattern analysis. ... (1) Imamura F, Jacques PF. Invited commentary: dietary pattern analysis. Am J Epidemiol 2011;173(0):000-000. (2) Gorst-Rasmussen A, Dahm CC, Dethlefsen C, et al. Exploring dietary patterns by using the treelet transform. Am J Epidemiol 2011;173(0):000-000
Analysis of spatial count data using Kalman smoothing
No full textThis paper considers spatial count data from an agricultural field experiment. Counts of weed plants in a field have been recorded in a project on precision farming. Interest is in mapping the weed intensity so that the dose of herbicide applied at any location can be adjusted to the amount of weed present at the location. We elaborate on a link between state space models and Markov random fields. The oberservations are modelled as independent Poisson counts conditional on a Gaussian Markov random field. We employ the fact that the model may be written as a state space model which may be analysed by combining approximate Kalman filter techniques with importance sampling
Analysis of spatial count data using Kalman smoothing
No full textWe consider spatial count data from an agricultural field experiment. Counts of weed plants in a field have been recorded in a project on precision farming. Interest is in mapping the weed intensity so that the dose of herbicide applied at any location can be adjusted to the amount of weed present at the location. We elaborate on a link between state space models and Markov random fields. The observations are modelled as independent Poisson counts conditional on a Gaussian Markov random field. We employ the fact that the model may be written as a state space model which may be analysed by combining approximate Kalman filter techniques with importance sampling.We consider spatial count data from an agricultural field experiment. Counts of weed plants in a field have been recorded in a project on precision farming. Interest is in mapping the weed intensity so that the dose of herbicide applied at any location can be adjusted to the amount of weed present at the location. We elaborate on a link between state space models and Markov random fields. The observations are modelled as independent Poisson counts conditional on a Gaussian Markov random field. We employ the fact that the model may be written as a state space model which may be analysed by combining approximate Kalman filter techniques with importance sampling.</p
Space time problems and applications
No full textState space models and Kalman filter techniques have been widely used for the analysis of time series. Typically, a latent process is assessed from observations using filtering (the present), smoothing (the past) and/or prediction (the future). The model class is very broad and comprises ARIMA models, cubic spline models and structural time series models. The development of state space theory has interacted with the development of other statistical disciplines. In the first part of the Thesis, we present the theory of state space models, including Gaussian state space models, approximative analysis of non-Gaussian models, simulation based techniques and model diagnostics. The second part of the Thesis considers Markov random field models. These are spatial models applicable in e.g. disease mapping and in agricultural experiments. Recently, the Gaussian Markov random field models were expressed as state space models, enabling the Kalman filter machinery. Our main contribution is to extend the Markov random field models by generalising the corresponding state space model. It turns out that several non-Gaussian spatial models can be analysed by combining approximate Kalman filter techniques with importance sampling. The third part of the Thesis contains applications of the theory. First, a univariate time series of count data is analysed. Then, a spatial model is used to compare wheat yields. Weed count data in connection with a project in precision farming is analysed using the developed methodology. Finally, a model for edge detection in digital images forms the basis of a simulation study
Hypothetical Estimands in Randomised Controlled Trials: Unifying Causal Inference and Semiparametric Theory
No full textWhen targeting the hypothetical estimand in a randomised controlled trial, accounting for intercurrent events in the analysis presents significant challenges as intercurrent events have a confounding effect. This project presents the causal inference workflow in the context of randomised clinical trials. In addition, the project presents the theory of semiparametric models in order to present the targeted learning framework. When determining the efficacy of treatments in terms of the hypothetical estimand, common practice is to use a Mixed Model for Repeated Measures (MMRM). This project proposes the use of Longitudinal Targeted Maximum Likelihood Estimation (LTMLE) for estimating the hypothetical estimand. Through simulations and empirical analysis, we assess how these methodologies manage the impact of varying amounts of intercurrent events on treatment outcomes. Our findings suggest that while MMRM provides an easily interpretable solution, LTMLE offers a more robust solution by more accurately reflecting the causal relationships in the presence of rescue medication and treatment discontinuation
- …
