We are delighted to announce that the esteemed speakers have graciously accepted our invitation to deliver keynote speeches at the Applied Statistics 2024 conference.
Emeritus Professor of Methodology and Statistics
University of Groningen, University of Oxford
Homepage
Abstract
Multilevel longitudinal network data sets are starting to be available more and more. This offers new possibilities for generalization of results of network analysis to populations of networks, and requires new methods of analysis. A multilevel version of the Stochastic Actor-Oriented Model (SAOM) was developed by Koskinen and Snijders (2023), and some experience with its use is presented here. In this hierarchical model, there are two sets of parameters: parameters at the highest level, the population of groups, which may be called the population parameters; and parameters at the group (or network) level, determining the network dynamics in each group, modelled by a SAOM. The distribution of the group-level parameters conditional on the population parameters is assumed to be multivariate normal. A fully Bayesian approach is followed in which the population-level parameters are treated as parameters with a prior distribution, and estimation is done by Markov chain Monte Carlo (MCMC) methods. To obtain good convergence of the MCMC algorithm, it is necessary to let only some of the group-level parameters vary across groups and keep the rest fixed, like in Hierarchical Linear Models for regular multilevel analysis. The analysis of each network then borrows strength from the data for the other networks, much like in the Hierarchical Linear Model. This method allows estimation of SAOMs for smaller data sets than is possible for single groups. An issue for the practical application is the necessity for the user to specify a prior distribution. Some examples are presented of the sensitivity of the estimation results to the specification of the prior.
Professor of Medical Statistics
Leiden University Medical Center
Homepage
Abstract
The Fine-Gray subdistribution hazard model has become the default method to estimate the incidence of outcomes over time in the presence of competing risks. This model is attractive because it directly relates covariates to the cumulative incidence function (CIF) of the event of interest. An alternative is to combine the different cause-specific hazard functions to obtain the different CIFs. A limitation of the subdistribution hazard approach is that the sum of the cause-specific CIFs can exceed 1 (100%) for some covariate patterns. Using data on 9479 patients hospitalized with acute myocardial infarction, we estimated the cumulative incidence of both cardiovascular death and non-cardiovascular death for each patient. We found that when using subdistribution hazard models, approximately 5% of subjects had an estimated risk of 5-year all-cause death (obtained by combining the two cause-specific CIFs obtained from subdistribution hazard models) that exceeded 1. This phenomenon was avoided by using the two cause-specific hazard models. We provide a proof that the sum of predictions exceeds 1 is a fundamental problem with the Fine-Gray subdistribution hazard model. Care should be taken when using the Fine-Gray subdistribution hazard model in situations with wide risk distributions or a high cumulative incidence, and if one is interested in the risk of failure from each of the different event types.