Prediction of Changes in Seafloor Depths Based on Time Series of Bathymetry Observations: Dutch North Sea Case

Guaranteeing safety of navigation within the Netherlands Continental Shelf (NCS), while efficiently using its ocean mapping resources, is a key task of Netherlands Hydrographic Service (NLHS) and Rijkswaterstaat (RWS). Resurvey frequencies depend on seafloor dynamics and the aim of this research is to model the seafloor dynamics to predict changes in seafloor depth that would require resurveying. Characterisation of the seafloor dynamics is based on available time series of bathymetry data obtained from the acoustic remote sensing method of both single-beam echosounding (SBES) and multibeam echosounding (MBES). This time series is used to define a library of mathematical models describing the seafloor dynamics in relation to spatial and temporal changes in depth. An adaptive, functional model selection procedure is developed using a nodal analysis (0D) approach, based on statistical hypothesis testing using a combination of the Overall Model Test (OMT) statistic and Generalised Likelihood Ratio Test (GLRT). This approach ensures that each model has an equal chance of being selected, when more than one hypothesis is plausible for areas that exhibit varying seafloor dynamics. This ensures a more flexible and rigorous decision on the choice of the nominal model assumption. The addition of piecewise linear models to the library offers another characterisation of the trends in the nodal time series. This has led to an optimised model selection procedure and parameterisation of each nodal time series, which is used for the spatial and temporal predictions of the changes in the depths and associated uncertainties. The model selection results show that the models can detect the changes in the seafloor depths with spatial consistency and similarity, particularly in the shoaling areas where tidal sandwaves are present. The predicted changes in depths and uncertainties are translated into a probability risk-alert map by evaluating the probabilities of an indicator variable exceeding a certain decision threshold. This research can further support the decision-making process when optimising resurvey frequencies.

Ambiguous support for extra-tropical accelerated diversification of rosids

Sun et al.1 used a comprehensive phylogenetic and a locality based climatic dataset to examine how past climates have driven diversification across 17 orders of rosids. They concluded that tropical rosids diversify slower than the (younger) non-tropical counterparts in part due to a strong negative relationship between paleo-temperature and diversification (i.e. higher diversification rates under lower temperatures). Their conclusions are based on tip rates (derived from BAMM2; DR3) and correlations with current temperature (e-SIM4) or binary tropicality data sets (BAMM/STRAPP5; FiSSE6) as well as tree-wide estimates of diversification with paleo-temperature (BAMM; RPANDA7) or tropicality (BiSSE8; HiSSE9). Here, I highlight several inconsistencies in their diversification analyses as well as a systematic error pertaining to the RPANDA model selection procedure, which, together with several minor technical issues, weaken the support for Sun et al.’s1 conclusions. A re-analysis of their BiSSE/HiSSE and RPANDA analyses are performed.

Prior hypotheses or regularization allow inference of diversification histories from extant timetrees

Phylogenies of extant species are widely used to study past diversification dynamics1. The most common approach is to formulate a set of candidate models representing evolutionary hypotheses for how and why speciation and extinction rates in a clade changed over time, and compare those models through their probability to have generated the corresponding empirical tree. Recently, Louca & Pennell2 reported the existence of an infinite number of ‘congruent’ models with potentially markedly different diversification dynamics, but equal likelihood, for any empirical tree (see also Lambert & Stadler3). Here we explore the implications of these results, and conclude that they neither undermine the hypothesis-driven model selection procedure widely used in the field nor show that speciation and extinction dynamics cannot be investigated from extant timetrees using a data-driven procedure.

Model Selection in Continuous Test Norming With GAMLSS

To compute norms from reference group test scores, continuous norming is preferred over traditional norming. A suitable continuous norming approach for continuous data is the use of the Box–Cox Power Exponential model, which is found in the generalized additive models for location, scale, and shape. Applying the Box–Cox Power Exponential model for test norming requires model selection, but it is unknown how well this can be done with an automatic selection procedure. In a simulation study, we compared the performance of two stepwise model selection procedures combined with four model-fit criteria (Akaike information criterion, Bayesian information criterion, generalized Akaike information criterion (3), cross-validation), varying data complexity, sampling design, and sample size in a fully crossed design. The new procedure combined with one of the generalized Akaike information criterion was the most efficient model selection procedure (i.e., required the smallest sample size). The advocated model selection procedure is illustrated with norming data of an intelligence test.

Bayesian analysis of fMRI data and RNA-Seq time course experiment data

The present dissertation contains two parts. In the first part, we develop a new Bayesian analysis of functional MRI data. We propose a novel triple gamma Hemodynamic Response Function (HRF) including the component to describe the initial dip. We use HRF to inform voxel-wise neuronal activities. Then we devise a new model selection procedure with a nonlocal pMOM prior for joint detection of neuronal activation and estimation of HRF, in order to time the activation time difference between visual and motor areas in the brain. In the second part, we develop a new Bayesian analysis of RNA-Seq Time Course experiments data. We propose to use Bayesian Principal Component regression model and based on that, devise a model selection procedure by using nonlocal piMOM prior in order to identify differentially expressed genes. Most current existing methods for RNA-Seq Time Course experiments data are from static view of point and cannot predict temporal patterns. Our method estimate the posterior differentially expressed probability for each gene by borrowing information across all subjects. Use of nonlocal prior in the model selection procedure reduces false discovered differentially expressed genes.