Evaluation of the accuracy of multilevel analysis parameter estimates for intra-cluster correlation in correlated binary data

Many studies in biostatistics deal with binary data. Some of these studies involve correlated observations, which can complicate the analysis of the resulting data. Studies of this kind typically arise when a high degree of commonality exists between test subjects. If there exists a natural hierarchy in the data, multilevel analysis is an appropriate tool for the analysis. Two examples are the measurements on identical twins, or the study of symmetrical organs or appendages such as in the case of ophthalmic studies. Although this type of matching appears ideal for the purposes of comparison, analysis of the resulting data while ignoring the effect of intra-cluster correlation has been shown to produce biased results.^ This paper will explore the use of multilevel modeling of simulated binary data with predetermined levels of correlation. Data will be generated using the Beta-Binomial method with varying degrees of correlation between the lower level observations. The data will be analyzed using the multilevel software package MlwiN (Woodhouse, et al, 1995). Comparisons between the specified intra-cluster correlation of these data and the estimated correlations, using multilevel analysis, will be used to examine the accuracy of this technique in analyzing this type of data. ^

Robust estimates of climate-induced hydrological change in a temperate mountainous region

A sustainable water resources management depends on sound information about the impacts of climate change. This information is, however, not easily derived because natural runoff variability interferes with the climate change signal. This study presents a procedure that leads to robust estimates of magnitude and Time Of Emergence (TOE) of climate-induced hydrological change that also account for the natural variability contained in the time series. Firstly, natural variability of 189 mesoscale catchments in Switzerland is sampled for 10 ENSEMBLES scenarios for the control (1984–2005) and two scenario periods (near future: 2025–2046, far future: 2074–2095) applying a bootstrap procedure. Then, the sampling distributions of mean monthly runoff are tested for significant differences with the Wilcoxon-Mann–Whitney test and for effect size with Cliff’s delta d. Finally, the TOE of a climate change induced hydrological change is determined when at least eight out of the ten hydrological projections significantly differ from natural variability. The results show that the TOE occurs in the near future period except for high-elevated catchments in late summer. The significant hydrological projections in the near future correspond, however, to only minor runoff changes. In the far future, hydrological change is statistically significant and runoff changes are substantial. Temperature change is the most important factor determining hydrological change in this mountainous region. Therefore, hydrological change depends strongly on a catchment’s mean elevation. Considering that the hydrological changes are predicted to be robust in the near future highlights the importance of accounting for these changes in water resources planning.

Model-Based Estimates of Calanus Finmarchicus Abundance in the Gulf of Maine

Ocean observing systems and satellites routinely collect a wealth of information on physical conditions in the ocean. With few exceptions, such as chlorophyll concentrations, information on biological properties is harder to measure autonomously. Here, we present a system to produce estimates of the distribution and abundance of the copepod Calanus finmarchicus in the Gulf of Maine. Our system uses satellite-based measurements of sea surface temperature and chlorophyll concentration to determine the developmental and reproductive rates of C. finmarchicus. The rate information then drives a population dynamics model of C. finmarchicus that is embedded in a 2-dimensional circulation field. The first generation of this system produces realistic information on interannual variability in C. finmarchicus distribution and abundance during the winter and spring. The model can also be used to identify key drivers of interannual variability in C. finmarchicus. Experiments with the model suggest that changes in initial conditions are overwhelmed by variability in growth rates after approximately 50 d. Temperature has the largest effect on growth rate. Elevated chlorophyll during the late winter can lead to increased C. finmarchicus abundance during the spring, but the effect of variations in chlorophyll concentrations is secondary to the other inputs. Our system could be used to provide real-time estimates or even forecasts of C. finmarchicus distribution. These estimates could then be used to support management of copepod predators such as herring and right whales.

Model-Based Estimates of Right Whale Habitat Use in the Gulf of Maine

Balancing human uses of the marine environment with the recovery of protected species requires accurate information on when and where species of interest are likely to be present. Here, we describe a system that can produce useful estimates of right whale Eubalaena glacialis presence and abundance on their feeding grounds in the Gulf of Maine. The foundation of our system is a coupled physical-biological model of the copepod Calan us finmarchicus, the preferred prey of right whales. From the modeled prey densities, we can estimate when whales will appear in the Great South Channel feeding ground. Based on our experience with the system, we consider how the relationship between right whales and copepods changes across spatial scales. The scale-dependent relationship between whales and copepods provides insight into how to improve future estimates of the distribution of right whales and other pelagic predators.

Experimenting and modeling: sharp distinctions, integration or just a mess?

At first sight, experimenting and modeling form two distinct modes of scientific inquiry. This spurs philosophical debates about how the distinction should be drawn (e.g. Morgan 2005, Winsberg 2009, Parker 2009). But much scientific practice casts serious doubts on the idea that the distinction makes much sense. There are two worries. First, the practices of modeling and experimenting are often intertwined in intricate ways because much modeling involves experimenting, and the interpretation of many experiments relies upon models. Second, there are borderline cases that seem to blur the distinction between experiment and model (if there is any). My talk tries to defend the philosophical project of distinguishing models from experiment and to advance the related philosophical debate. I begin with providing a minimalist framework of conceptualizing experimenting and modeling and their mutual relationships. The methods are conceptualized as different types of activities that are characterized by a primary goal, respectively. The minimalist framwork, which should be uncontroversial, suffices to accommodate the first worry. I address the second worry by suggesting several ways how to conceptualize the distinction in a more flexible way. I make a concrete suggestion of how the distinction may be drawn. I use examples from the history of science to argue my case. The talk concentrates and models and experiments, but I will comment on simulations too.

Plotting regression coefficients and other estimates

Graphical display of regression results has become increasingly popular in presentations and in scientific literature because graphs are often much easier to read than tables. Such plots can be produced in Stata by the marginsplot command (see [R] marginsplot). However, while marginsplot is versatile and flexible, it has two major limitations: it can only process results left behind by margins (see [R] margins), and it can handle only one set of results at a time. In this article, I introduce a new command called coefplot that overcomes these limitations. It plots results from any estimation command and combines results from several models into one graph. The default behavior of coefplot is to plot markers for coefficients and horizontal spikes for confidence intervals. However, coefplot can also produce other types of graphs. I illustrate the capabilities of coefplot by using a series of examples.

Sharp neck injuries in suicidal intention

Sharp neck injuries in suicidal intention often present as serious emergency situations with the need for an immediate diagnosis and treatment. We report our study of the clinical evolution of this emergency condition. This study investigates the cases of sharp neck injuries in suicidal intention treated at our institution between 2000 and 2010. Patient records were collected in a retrospectively reviewed and analyzed database. The current literature was compared to our findings. We found 36 cases (10 female and 26 male). The neck injuries were superficial and profound in 16 and 20 patients, respectively. Twenty-two patients were seen by the Head and Neck surgeon. A surgical neck exploration was necessary in 19 cases. Tracheal, laryngeal, pharyngeal and vascular injuries were found in one, five, three and three cases, respectively. The hospital stay ranged from 1 to 47 days. All the patients underwent emergency psychiatric assessment and were subsequently referred for psychiatric treatment. One patient died in the emergency room from an additional arterial injury to the wrist. Sharp neck injuries in suicidal intention treated with an interdisciplinary medical, surgical and psychiatric emergency assessment and treatment have low mortality and morbidity.

Making regression tables from stored estimates

Organizing and archiving statistical results and processing a subset of those results for publication are important and often underestimated issues in conducting statistical analyses. Because automation of these tasks is often poor, processing results produced by statistical packages is quite laborious and vulnerable to error. I will therefore present a new package called estout that facilitates and automates some of these tasks. This new command can be used to produce regression tables for use with spreadsheets, LaTeX, HTML, or word processors. For example, the results for multiple models can be organized in spreadsheets and can thus be archived in an orderly manner. Alternatively, the results can be directly saved as a publication-ready table for inclusion in, for example, a LaTeX document. estout is implemented as a wrapper for estimates table but has many additional features, such as support for mfx. However, despite its flexibility, estout is—I believe—still very straightforward and easy to use. Furthermore, estout can be customized via so-called defaults files. A tool to make available supplementary statistics called estadd is also provided.

Time Variable Gravity: Contributions of GOCE Satellite Data to Monthly and Bi-monthly GRACE Gravity Estimates

A feasibility study by Pail et al. (Can GOCE help to improve temporal gravity field estimates? In: Ouwehand L (ed) Proceedings of the 4th International GOCE User Workshop, ESA Publication SP-696, 2011b) shows that GOCE (‘Gravity field and steady-state Ocean Circulation Explorer’) satellite gravity gradiometer (SGG) data in combination with GPS derived orbit data (satellite-to-satellite tracking: SST-hl) can be used to stabilize and reduce the striping pattern of a bi-monthly GRACE (‘Gravity Recovery and Climate Experiment’) gravity field estimate. In this study several monthly (and bi-monthly) combinations of GRACE with GOCE SGG and GOCE SST-hl data on the basis of normal equations are investigated. Our aim is to assess the role of the gradients (solely) in the combination and whether already one month of GOCE observations provides sufficient data for having an impact in the combination. The estimation of clean and stable monthly GOCE SGG normal equations at high resolution ( >  d/o 150) is found to be difficult, and the SGG component, solely, does not show significant added value to monthly and bi-monthly GRACE gravity fields. Comparisons of GRACE-only and combined monthly and bi-monthly solutions show that the striping pattern can only be reduced when using both GOCE observation types (SGG, SST-hl), and mainly between d/o 45 and 60.

Pollen dispersal properties of Poaceae and Cyperaceae: First estimates of their absolute pollen productivities

Pollen-trap results from the Swiss Alps 1996–2009 were used to assess the pollen dispersal–deposition properties of Poaceae (grasses) and Cyperaceae (sedges). Dispersal parameter values were investigated for a modified version of the Prentice–Sugita pollen dispersal–deposition model. Appropriate values (i.e. realistic in the field and allowing realistic modelling results) for wind speed are suggested to be in the range of 3–7 m s− 1 and for pollen an injection height of 0.03–0.1 m above the ground. The appropriate range of pollen injection height values for grasses and sedges differs from that of trees in the same area, suggesting different pollen dispersal properties between herbs and trees. In addition, logarithmic weighting of the vegetation was tested as an alternative to the modified Prentice–Sugita model. This yielded very similar results, suggesting that the use of such much simpler approximations of the pollen–vegetation relationship is a plausible alternative. Based on the modified Prentice–Sugita model, absolute pollen productivity for Poaceae was estimated to 7300 ± 400 grains cm− 2 year− 1 (1 SE). The data basis for Cyperaceae is smaller than for Poaceae, but the dispersal parameter values determined as appropriate for Poaceae yield good results. Absolute pollen productivity for Cyperaceae was estimated to 6300 ± 1100 grains cm− 2 year− 1 (1 SE).

Ignoring heterozygous sites biases phylogenomic estimates of divergence times: implications for the evolutionary history of Microtus voles

Phylogenetic reconstruction of the evolutionary history of closely related organisms may be difficult because of the presence of unsorted lineages and of a relatively high proportion of heterozygous sites that are usually not handled well by phylogenetic programs. Genomic data may provide enough fixed polymorphisms to resolve phylogenetic trees, but the diploid nature of sequence data remains analytically challenging. Here, we performed a phylogenomic reconstruction of the evolutionary history of the common vole (Microtus arvalis) with a focus on the influence of heterozygosity on the estimation of intraspecific divergence times. We used genome-wide sequence information from 15 voles distributed across the European range. We provide a novel approach to integrate heterozygous information in existing phylogenetic programs by repeated random haplotype sampling from sequences with multiple unphased heterozygous sites. We evaluated the impact of the use of full, partial, or no heterozygous information for tree reconstructions on divergence time estimates. All results consistently showed four deep and strongly supported evolutionary lineages in the vole data. These lineages undergoing divergence processes split only at the end or after the last glacial maximum based on calibration with radiocarbon-dated paleontological material. However, the incorporation of information from heterozygous sites had a significant impact on absolute and relative branch length estimations. Ignoring heterozygous information led to an overestimation of divergence times between the evolutionary lineages of M. arvalis. We conclude that the exclusion of heterozygous sites from evolutionary analyses may cause biased and misleading divergence time estimates in closely related taxa.

Quantifying differences in land use emission estimates implied by definition discrepancies

The quantification of CO2 emissions from anthropogenic land use and land use change (eLUC) is essential to understand the drivers of the atmospheric CO2 increase and to inform climate change mitigation policy. Reported values in synthesis reports are commonly derived from different approaches (observation-driven bookkeeping and process-modelling) but recent work has emphasized that inconsistencies between methods may imply substantial differences in eLUC estimates. However, a consistent quantification is lacking and no concise modelling protocol for the separation of primary and secondary components of eLUC has been established. Here, we review differences of eLUC quantification methods and apply an Earth System Model (ESM) of Intermediate Complexity to quantify them. We find that the magnitude of effects due to merely conceptual differences between ESM and offline vegetation model-based quantifications is ~ 20 % for today. Under a future business-as-usual scenario, differences tend to increase further due to slowing land conversion rates and an increasing impact of altered environmental conditions on land-atmosphere fluxes. We establish how coupled Earth System Models may be applied to separate secondary component fluxes of eLUC arising from the replacement of potential C sinks/sources and the land use feedback and show that secondary fluxes derived from offline vegetation models are conceptually and quantitatively not identical to either, nor their sum. Therefore, we argue that synthesis studies should resort to the "least common denominator" of different methods, following the bookkeeping approach where only primary land use emissions are quantified under the assumption of constant environmental boundary conditions.