The Potential of Subsampling and Inpainting for Fast Low-Dose Cryo FIB-SEM Imaging

Traditional image acquisition for cryo focused ion-beam scanning electron microscopy (FIB-SEM) tomography often sees thousands of images being captured over a period of many hours, with immense data sets being produced. When imaging beam sensitive materials, these images are often compromised by additional constraints related to beam damage and the devitrification of the material during imaging, which renders data acquisition both costly and unreliable. Subsampling and inpainting are proposed as solutions for both of these aspects, allowing fast and low-dose imaging to take place in the Focused ion-beam scanning electron microscopy FIB-SEM without an appreciable loss in image quality. In this work, experimental data are presented which validate subsampling and inpainting as a useful tool for convenient and reliable data acquisition in a FIB-SEM, with new methods of handling three-dimensional data being employed in the context of dictionary learning and inpainting algorithms using a newly developed microscope control software and data recovery algorithm.

The recovery of parabolic avalanches in spatially subsampled neuronal networks at criticality

Scaling relationships characterize complex systems at criticality. In the brain, these relationships are evident in scale-invariant activity cascades, so-called neuronal avalanches, quantified by power laws in avalanche size and duration. At the cellular level, neuronal avalanches are identified in spatially distributed groups of neurons that participate in cascades of coincident action potential firing. Such spatiotemporal synchronization is central to theories on brain function, yet scaling relationships in avalanche synchronization have been challenging to study when only a fraction of neurons is observed, underestimating avalanche properties. Here, we study these biases from fractional sampling in an all-to-all, balanced network of excitatory and inhibitory neurons with critical branching process dynamics. We focus on the growth of mean avalanche size with avalanche duration. For parabolic avalanches, this growth is quadratic, quantified by the scaling exponent, χ = 2 , which signifies rapid spatial expansion of coincident firing within a relatively short period of time. In contrast, χ < < 2 for fractionally sampled networks. We show that temporal coarse-graining combined with a threshold for the minimally required coincident firing in the network recovers χ = 2 , even when sampling as few as 0.1% of the neurons. In contrast, a commonly proposed 'crackling noise' approach fails to recover χ under those conditions. Our approach robustly identifies χ = 2 for ongoing neuronal activity in frontal cortex of awake mice using cellular 2-photon imaging. Our findings demonstrate how to correct scaling bias from fractional sampling and identifies rapid, scale-invariant synchronization of cell assemblies in the brain.

An examination of model fit and measurement invariance of general mental ability and personality measures used in the multilingual context of the Swiss Armed Forces: A Bayesian structural equation modeling approach

Measurement invariance of psychological test batteries is an essential quality criterion when the test batteries are administered in different cultural and language contexts. The purpose of this study was to examine to what extent measurement model fit and measurement invariance across the two largest language groups in Switzerland (i.e., German and French speakers) can be assumed for selected general mental ability and personality tests used in the Swiss Armed Forces' cadre selection process. For the model fit and invariance testing, we used Bayesian structural equation modeling (BSEM). Because the sizes of the language group samples were unbalanced, we reran the invariance testing with the subsampling procedure as a robustness check. The results showed that at least partial approximate scalar invariance can be assumed for the constructs. However, comparisons in the full sample and subsamples also showed that certain test items function differently across the language groups. The results are discussed regarding the three following issues: First, we critically discuss the applied criterion and alternative effect size measures for assessing the practical importance of non-invariances. Second, we highlight potential remedies and further testing options, that can be applied, once certain items have been detected to function differently. Third, we discuss alternative modeling and invariance testing approaches to BSEM and outline future research avenues.

Mapping the Dynamics of Contemporary PRRSV-2 Evolution and Its Emergence and Spreading Hotspots in the U.S. Using Phylogeography

The repeated emergence of new genetic variants of PRRSV-2, the virus that causes porcine reproductive and respiratory syndrome (PRRS), reflects its rapid evolution and the failure of previous control efforts. Understanding spatiotemporal heterogeneity in variant emergence and spread is critical for future outbreak prevention. Here, we investigate how the pace of evolution varies across time and space, identify the origins of sub-lineage emergence, and map the patterns of the inter-regional spread of PRRSV-2 Lineage 1 (L1)-the current dominant lineage in the U.S. We performed comparative phylogeographic analyses on subsets of 19,395 viral ORF5 sequences collected across the U.S. and Canada between 1991 and 2021. The discrete trait analysis of multiple spatiotemporally stratified sampled sets (n = 500 each) was used to infer the ancestral geographic region and dispersion of each sub-lineage. The robustness of the results was compared to that of other modeling methods and subsampling strategies. Generally, the spatial spread and population dynamics varied across sub-lineages, time, and space. The Upper Midwest was a main spreading hotspot for multiple sub-lineages, e.g., L1C and L1F, though one of the most recent emergence events (L1A(2)) spread outwards from the east. An understanding of historical patterns of emergence and spread can be used to strategize disease control and the containment of emerging variants.

Towards real-time STEM simulations through targeted subsampling strategies

Scanning transmission electron microscopy images can be complex to interpret on the atomic scale as the contrast is sensitive to multiple factors such as sample thickness, composition, defects and aberrations. Simulations are commonly used to validate or interpret real experimental images, but they come at a cost of either long computation times or specialist hardware such as graphics processing units. Recent works in compressive sensing for experimental STEM images have shown that it is possible to significantly reduce the amount of acquired signal and still recover the full image without significant loss of image quality, and therefore it is proposed here that similar methods can be applied to STEM simulations. In this paper, we demonstrate a method that can significantly increase the efficiency of STEM simulations through a targeted sampling strategy, along with a new approach to independently subsample each frozen phonon layer. We show the effectiveness of this method by simulating a SrTiO₃ grain boundary and monolayer 2H-MoS₂ containing a sulphur vacancy using the abTEM software. We also show how this method is not limited to only traditional multislice methods, but also increases the speed of the PRISM simulation method. Furthermore, we discuss the possibility for STEM simulations to seed the acquisition of real data, to potentially lead the way to self-driving (correcting) STEM.

Understanding aliasing effects and their removal in SPEN MRI: A k-space perspective

PURPOSE

To characterize the mechanism of formation and the removal of aliasing artifacts and edge ghosts in spatiotemporally encoded (SPEN) MRI within a k-space theoretical framework.

METHODS

SPEN's quadratic phase modulation can be described in k-space by a convolution matrix whose coefficients derive from Fourier relations. This k-space model allows us to pose SPEN's reconstruction as a deconvolution process from which aliasing and edge ghost artifacts can be quantified by estimating the difference between a full sampling and reconstructions resulting from undersampled SPEN data.

RESULTS

Aliasing artifacts in SPEN MRI reconstructions can be traced to image contributions corresponding to high-frequency k-space signals. The k-space picture provides the spatial displacements, phase offsets, and linear amplitude modulations associated to these artifacts, as well as routes to removing these from the reconstruction results. These new ways to estimate the artifact priors were applied to reduce SPEN reconstruction artifacts on simulated, phantom, and human brain MRI data.

CONCLUSION

A k-space description of SPEN's reconstruction helps to better understand the signal characteristics of this MRI technique, and to improve the quality of its resulting images.

Analog Lock-In Amplifier Design Using Subsampling for Accuracy Enhancement in GMI Sensor Applications

A frequency downscaling technique for enhancing the accuracy of analog lock-in amplifier (LIA) architectures in giant magneto-impedance (GMI) sensor applications is presented in this paper. As a proof of concept, the proposed method is applied to two different LIA topologies using, respectively, analog and switching-based multiplication for phase-sensitive detection. Specifically, the operation frequency of both the input and the reference signals of the phase-sensitive detector (PSD) block of the LIA is reduced through a subsampling process using sample-and-hold (SH) circuits. A frequency downscaling from 200 kHz, which is the optimal operating frequency of the employed GMI sensor, to 1 kHz has been performed. In this way, the proposed technique exploits the inherent advantages of analog signal multiplication at low frequencies, while the principle of operation of the PSD remains unaltered. The circuits were assembled using discrete components, and the frequency downscaling proposal was experimentally validated by comparing the measurement accuracy with the equivalent conventional circuits. The experimental results revealed that the error in the signal magnitude measurements was reduced by a factor of 8 in the case of the analog multipliers and by a factor of 21 when a PSD based on switched multipliers was used. The error in-phase detection using a two-phase LIA was also reduced by more than 25%.

How to remove or control confounds in predictive models, with applications to brain biomarkers

BACKGROUND

With increasing data sizes and more easily available computational methods, neurosciences rely more and more on predictive modeling with machine learning, e.g., to extract disease biomarkers. Yet, a successful prediction may capture a confounding effect correlated with the outcome instead of brain features specific to the outcome of interest. For instance, because patients tend to move more in the scanner than controls, imaging biomarkers of a disease condition may mostly reflect head motion, leading to inefficient use of resources and wrong interpretation of the biomarkers.

RESULTS

Here we study how to adapt statistical methods that control for confounds to predictive modeling settings. We review how to train predictors that are not driven by such spurious effects. We also show how to measure the unbiased predictive accuracy of these biomarkers, based on a confounded dataset. For this purpose, cross-validation must be modified to account for the nuisance effect. To guide understanding and practical recommendations, we apply various strategies to assess predictive models in the presence of confounds on simulated data and population brain imaging settings. Theoretical and empirical studies show that deconfounding should not be applied to the train and test data jointly: modeling the effect of confounds, on the training data only, should instead be decoupled from removing confounds.

CONCLUSIONS

Cross-validation that isolates nuisance effects gives an additional piece of information: confound-free prediction accuracy.

Reduction of Survey Sites in Dialectology: A New Methodology Based on Clustering

Many language change studies aim for a partial revisitation, i.e., selecting survey sites from previous dialect studies. The central issue of survey site reduction, however, has often been addressed only qualitatively. Cluster analysis offers an innovative means of identifying the most representative survey sites among a set of original survey sites. In this paper, we present a general methodology for finding representative sites for an intended study, potentially applicable to any collection of data about dialects or linguistic variation. We elaborate the quantitative steps of the proposed methodology in the context of the "Linguistic Atlas of Japan" (LAJ). Next, we demonstrate the full application of the methodology on the "Linguistic Atlas of German-speaking Switzerland" (Germ.: "Sprachatlas der Deutschen Schweiz"-SDS), with the explicit aim of selecting survey sites corresponding to the aims of the current project "Swiss German Dialects Across Time and Space" (SDATS), which revisits SDS 70 years later. We find that depending on the circumstances and requirements of a study, the proposed methodology, introducing cluster analysis into the survey site reduction process, allows for a greater objectivity in comparison to traditional approaches. We suggest, however, that the suitability of any set of candidate survey sites resulting from the proposed methodology be rigorously revised by experts due to potential incongruences, such as the overlap of objectives and variables across the original and intended studies and ongoing dialect change.

Subsampled Directed-Percolation Models Explain Scaling Relations Experimentally Observed in the Brain

Recent experimental results on spike avalanches measured in the urethane-anesthetized rat cortex have revealed scaling relations that indicate a phase transition at a specific level of cortical firing rate variability. The scaling relations point to critical exponents whose values differ from those of a branching process, which has been the canonical model employed to understand brain criticality. This suggested that a different model, with a different phase transition, might be required to explain the data. Here we show that this is not necessarily the case. By employing two different models belonging to the same universality class as the branching process (mean-field directed percolation) and treating the simulation data exactly like experimental data, we reproduce most of the experimental results. We find that subsampling the model and adjusting the time bin used to define avalanches (as done with experimental data) are sufficient ingredients to change the apparent exponents of the critical point. Moreover, experimental data is only reproduced within a very narrow range in parameter space around the phase transition.

Selection of variables for multivariable models: Opportunities and limitations in quantifying model stability by resampling

Statistical models are often fitted to obtain a concise description of the association of an outcome variable with some covariates. Even if background knowledge is available to guide preselection of covariates, stepwise variable selection is commonly applied to remove irrelevant ones. This practice may introduce additional variability and selection is rarely certain. However, these issues are often ignored and model stability is not questioned. Several resampling-based measures were proposed to describe model stability, including variable inclusion frequencies (VIFs), model selection frequencies, relative conditional bias (RCB), and root mean squared difference ratio (RMSDR). The latter two were recently proposed to assess bias and variance inflation induced by variable selection. Here, we study the consistency and accuracy of resampling estimates of these measures and the optimal choice of the resampling technique. In particular, we compare subsampling and bootstrapping for assessing stability of linear, logistic, and Cox models obtained by backward elimination in a simulation study. Moreover, we exemplify the estimation and interpretation of all suggested measures in a study on cardiovascular risk. The VIF and the model selection frequency are only consistently estimated in the subsampling approach. By contrast, the bootstrap is advantageous in terms of bias and precision for estimating the RCB as well as the RMSDR. Though, unbiased estimation of the latter quantity requires independence of covariates, which is rarely encountered in practice. Our study stresses the importance of addressing model stability after variable selection and shows how to cope with it.

PIQMEE: Bayesian Phylodynamic Method for Analysis of Large Data Sets with Duplicate Sequences

Next-generation sequencing of pathogen quasispecies within a host yields data sets of tens to hundreds of unique sequences. However, the full data set often contains thousands of sequences, because many of those unique sequences have multiple identical copies. Data sets of this size represent a computational challenge for currently available Bayesian phylogenetic and phylodynamic methods. Through simulations, we explore how large data sets with duplicate sequences affect the speed and accuracy of phylogenetic and phylodynamic analysis within BEAST 2. We show that using unique sequences only leads to biases, and using a random subset of sequences yields imprecise parameter estimates. To overcome these shortcomings, we introduce PIQMEE, a BEAST 2 add-on that produces reliable parameter estimates from full data sets with increased computational efficiency as compared with the currently available methods within BEAST 2. The principle behind PIQMEE is to resolve the tree structure of the unique sequences only, while simultaneously estimating the branching times of the duplicate sequences. Distinguishing between unique and duplicate sequences allows our method to perform well even for very large data sets. Although the classic method converges poorly for data sets of 6,000 sequences when allowed to run for 7 days, our method converges in slightly more than 1 day. In fact, PIQMEE can handle data sets of around 21,000 sequences with 20 unique sequences in 14 days. Finally, we apply the method to a real, within-host HIV sequencing data set with several thousand sequences per patient.

Towards Raman Automation for Microplastics: Developing Strategies for Particle Adhesion and Filter Subsampling

Automation and subsampling have been proposed as solutions to reduce the time required to quantify and characterize microplastics in samples using spectroscopy. However, there are methodological dilemmas associated with automation that are preventing its widespread implementation including ensuring particles stay adhered to the filter during filter mapping and developing an appropriate subsampling strategy to reduce the time needed for analysis. We provide a solution to the particle adherence issue by applying Skin Tac, a non-polymeric permeable adhesive that allows microplastic particles to adhere to the filter without having their Raman signal masked by the adhesive. We also explore different subsampling strategies to help inform how to take a representative subsample. Based on the particle distributions observed on filters, we determined that assuming a homogenous particle distribution is inappropriate and can lead to over- and under-estimations of extrapolated particle counts. Instead, we provide recommendations for future studies that wish to subsample to increase the throughput of samples for spectroscopic analysis.

Stable polyp-scene classification via subsampling and residual learning from an imbalanced large dataset

This Letter presents a stable polyp-scene classification method with low false positive (FP) detection. Precise automated polyp detection during colonoscopies is essential for preventing colon-cancer deaths. There is, therefore, a demand for a computer-assisted diagnosis (CAD) system for colonoscopies to assist colonoscopists. A high-performance CAD system with spatiotemporal feature extraction via a three-dimensional convolutional neural network (3D CNN) with a limited dataset achieved about 80% detection accuracy in actual colonoscopic videos. Consequently, further improvement of a 3D CNN with larger training data is feasible. However, the ratio between polyp and non-polyp scenes is quite imbalanced in a large colonoscopic video dataset. This imbalance leads to unstable polyp detection. To circumvent this, the authors propose an efficient and balanced learning technique for deep residual learning. The authors’ method randomly selects a subset of non-polyp scenes whose number is the same number of still images of polyp scenes at the beginning of each epoch of learning. Furthermore, they introduce post-processing for stable polyp-scene classification. This post-processing reduces the FPs that occur in the practical application of polyp-scene classification. They evaluate several residual networks with a large polyp-detection dataset consisting of 1027 colonoscopic videos. In the scene-level evaluation, their proposed method achieves stable polyp-scene classification with 0.86 sensitivity and 0.97 specificity.

Likelihood Inference for Large Scale Stochastic Blockmodels with Covariates based on a Divide-and-Conquer Parallelizable Algorithm with Communication

We consider a stochastic blockmodel equipped with node covariate information, that is helpful in analyzing social network data. The key objective is to obtain maximum likelihood estimates of the model parameters. For this task, we devise a fast, scalable Monte Carlo EM type algorithm based on case-control approximation of the log-likelihood coupled with a subsampling approach. A key feature of the proposed algorithm is its parallelizability, by processing portions of the data on several cores, while leveraging communication of key statistics across the cores during each iteration of the algorithm. The performance of the algorithm is evaluated on synthetic data sets and compared with competing methods for blockmodel parameter estimation. We also illustrate the model on data from a Facebook derived social network enhanced with node covariate information.

On the Sensitivity of Granger Causality to Errors-In-Variables, Linear Transformations and Subsampling

This article studies the sensitivity of Granger causality to the addition of noise, the introduction of subsampling, and the application of causal invertible filters to weakly stationary processes. Using canonical spectral factors and Wold decompositions, we give general conditions under which additive noise or filtering distorts Granger-causal properties by inducing (spurious) Granger causality, as well as conditions under which it does not. For the errors-in-variables case, we give a continuity result, which implies that: a 'small' noise-to-signal ratio entails 'small' distortions in Granger causality. On filtering, we give general necessary and sufficient conditions under which 'spurious' causal relations between (vector) time series are not induced by linear transformations of the variables involved. This also yields transformations (or filters) which can eliminate Granger causality from one vector to another one. In a number of cases, we clarify results in the existing literature, with a number of calculations streamlining some existing approaches.

Computationally efficient methods for fitting mixed models to electronic health records data

Motivated by two case studies using primary care records from the Clinical Practice Research Datalink, we describe statistical methods that facilitate the analysis of tall data, with very large numbers of observations. Our focus is on investigating the association between patient characteristics and an outcome of interest, while allowing for variation among general practices. We explore ways to fit mixed-effects models to tall data, including predictors of interest and confounding factors as covariates, and including random intercepts to allow for heterogeneity in outcome among practices. We introduce (1) weighted regression and (2) meta-analysis of estimated regression coefficients from each practice. Both methods reduce the size of the dataset, thus decreasing the time required for statistical analysis. We compare the methods to an existing subsampling approach. All methods give similar point estimates, and weighted regression and meta-analysis give similar standard errors for point estimates to analysis of the entire dataset, but the subsampling method gives larger standard errors. Where all data are discrete, weighted regression is equivalent to fitting the mixed model to the entire dataset. In the presence of a continuous covariate, meta-analysis is useful. Both methods are easy to implement in standard statistical software.

Standard errors and confidence intervals for variable importance in random forest regression, classification, and survival

Random forests are a popular nonparametric tree ensemble procedure with broad applications to data analysis. While its widespread popularity stems from its prediction performance, an equally important feature is that it provides a fully nonparametric measure of variable importance (VIMP). A current limitation of VIMP, however, is that no systematic method exists for estimating its variance. As a solution, we propose a subsampling approach that can be used to estimate the variance of VIMP and for constructing confidence intervals. The method is general enough that it can be applied to many useful settings, including regression, classification, and survival problems. Using extensive simulations, we demonstrate the effectiveness of the subsampling estimator and in particular find that the delete-d jackknife variance estimator, a close cousin, is especially effective under low subsampling rates due to its bias correction properties. These 2 estimators are highly competitive when compared with the .164 bootstrap estimator, a modified bootstrap procedure designed to deal with ties in out-of-sample data. Most importantly, subsampling is computationally fast, thus making it especially attractive for big data settings.

Lepidosaurian diversity in the Mesozoic-Palaeogene: the potential roles of sampling biases and environmental drivers

Lepidosauria is a speciose clade with a long evolutionary history, but there have been few attempts to explore its taxon richness through time. Here we estimate patterns of terrestrial lepidosaur genus diversity for the Triassic-Palaeogene (252-23 Ma), and compare observed and sampling-corrected richness curves generated using Shareholder Quorum Subsampling and classical rarefaction. Generalized least-squares regression (GLS) is used to investigate the relationships between richness, sampling and environmental proxies. We found low levels of richness from the Triassic until the Late Cretaceous (except in the Kimmeridgian-Tithonian of Europe). High richness is recovered for the Late Cretaceous of North America, which declined across the K-Pg boundary but remained relatively high throughout the Palaeogene. Richness decreased following the Eocene-Oligocene Grande Coupure in North America and Europe, but remained high in North America and very high in Europe compared to the Late Cretaceous; elsewhere data are lacking. GLS analyses indicate that sampling biases (particularly, the number of fossil collections per interval) are the best explanation for long-term face-value genus richness trends. The lepidosaur fossil record presents many problems when attempting to reconstruct past diversity, with geographical sampling biases being of particular concern, especially in the Southern Hemisphere.

Spatiotemporal Pixelization to Increase the Recognition Score of Characters for Retinal Prostheses

Most of the retinal prostheses use a head-fixed camera and a video processing unit. Some studies proposed various image processing methods to improve visual perception for patients. However, previous studies only focused on using spatial information. The present study proposes a spatiotemporal pixelization method mimicking fixational eye movements to generate stimulation images for artificial retina arrays by combining spatial and temporal information. Input images were sampled with a resolution that was four times higher than the number of pixel arrays. We subsampled this image and generated four different phosphene images. We then evaluated the recognition scores of characters by sequentially presenting phosphene images with varying pixel array sizes (6 × 6, 8 × 8 and 10 × 10) and stimulus frame rates (10 Hz, 15 Hz, 20 Hz, 30 Hz, and 60 Hz). The proposed method showed the highest recognition score at a stimulus frame rate of approximately 20 Hz. The method also significantly improved the recognition score for complex characters. This method provides a new way to increase practical resolution over restricted spatial resolution by merging the higher resolution image into high-frame time slots.

Quantifying the responses of biological indices to rare macroinvertebrate taxa exclusion: Does excluding more rare taxa cause more error?

Including or excluding rare taxa in bioassessment is a controversial topic, which essentially affects the reliability and accuracy of the result. In the present paper, we hypothesize that biological indices such as Shannon-Wiener index, Simpson's index, Margalef index, evenness, BMWP (biological monitoring working party), and ASPT (Average Score Per Taxon) respond differently to rare taxa exclusion. To test this hypothesis, a benthic macroinvertebrate data set based on recent fifteen-year studies in China was built for suppositional plot analyses. A field research was conducted in the Nansi Lake to perform related analyses. The results of suppositional plot simulations showed that Simpson's index placed more weight on common taxa than any other studied indices, followed by Shannon-Wiener index which remained a high value with the exclusion of rare taxa. The results indicated that there was not much of effect on Simpson's index and Shannon-Wiener index when rare taxa were excluded. Rare taxa played an important role in Margalef index and BMWP than in other indices. Evenness showed an increase trend, while ASPT varied inconsistently with the exclusion of rare taxa. Results of the field study also indicated that rare taxa had few impacts on the Shannon-Wiener index. By examining the relationships between the rare taxa and biological indices in our study, it is suggested that including the rare taxa when using BMWP and excluding them in the proposed way (e.g., fixed-count subsampling) to calculate Shannon-Wiener index and Simpson's index could raise the efficiency and reduce the biases in the bioassessment of freshwater ecosystems.

Latitudinal diversity gradients in Mesozoic non-marine turtles

The latitudinal biodiversity gradient (LBG)-the pattern of increasing taxonomic richness with decreasing latitude-is prevalent in the structure of the modern biota. However, some freshwater taxa show peak richness at mid-latitudes; for example, extant Testudines (turtles, terrapins and tortoises) exhibit their greatest diversity at 25° N, a pattern sometimes attributed to recent bursts of climatically mediated species diversification. Here, we test whether this pattern also characterizes the Mesozoic distribution of turtles, to determine whether it was established during either their initial diversification or as a more modern phenomenon. Using global occurrence data for non-marine testudinate genera, we find that subsampled richness peaks at palaeolatitudes of 15-30° N in the Jurassic, 30-45° N through the Cretaceous to the Campanian, and from 30° to 60° N in the Maastrichtian. The absence of a significant diversity peak in southern latitudes is consistent with results from climatic models and turtle niche modelling that demonstrate a dearth of suitable turtle habitat in Gondwana during the Jurassic and Late Cretaceous. Our analyses confirm that the modern testudinate LBG has a deep-time origin and further demonstrate that LBGs are not always expressed as a smooth, equator-to-pole distribution.

LOCAL CASE-CONTROL SAMPLING: EFFICIENT SUBSAMPLING IN IMBALANCED DATA SETS

For classification problems with significant class imbalance, subsampling can reduce computational costs at the price of inflated variance in estimating model parameters. We propose a method for subsampling efficiently for logistic regression by adjusting the class balance locally in feature space via an accept-reject scheme. Our method generalizes standard case-control sampling, using a pilot estimate to preferentially select examples whose responses are conditionally rare given their features. The biased subsampling is corrected by a post-hoc analytic adjustment to the parameters. The method is simple and requires one parallelizable scan over the full data set. Standard case-control sampling is inconsistent under model misspecification for the population risk-minimizing coefficients θ*. By contrast, our estimator is consistent for θ* provided that the pilot estimate is. Moreover, under correct specification and with a consistent, independent pilot estimate, our estimator has exactly twice the asymptotic variance of the full-sample MLE-even if the selected subsample comprises a miniscule fraction of the full data set, as happens when the original data are severely imbalanced. The factor of two improves to [Formula: see text] if we multiply the baseline acceptance probabilities by c > 1 (and weight points with acceptance probability greater than 1), taking roughly [Formula: see text] times as many data points into the subsample. Experiments on simulated and real data show that our method can substantially outperform standard case-control subsampling.

Long-term trends mask variation in the direction and magnitude of short-term phenological shifts

PREMISE OF THE STUDY

Plants are flowering earlier in response to climate change. However, substantial interannual variation in phenology may make it difficult to discern and compare long-term trends. In addition to providing insight on data requirements for discerning such trends, phenological shifts within subsets of long-term records will provide insight into the mechanisms driving changes in flowering over longer time scales.

METHODS

To examine variation in flowering shifts among temporal subsets of long-term records, we used two data sets of flowering phenology from snow-dominated habitats: subalpine meadow in Gothic, Colorado, USA (38 yr), and arctic tundra in Zackenberg, Greenland (16 yr). Shifts in flowering time were calculated as 10-yr moving averages for onset, peak, and end of flowering.

KEY RESULTS

Flowering advanced over the course of the entire time series at both sites. Flowering shifts at Gothic were variable, with some 10-yr time frames showing significant delays and others significant advancements. Early-flowering species were more responsive than later-flowering species, while the opposite was true at Zackenberg. Flowering shifts at Zackenberg were less variable, with advanced flowering across all 10-yr time frames. At both sites, long-term advancement seemed to be primarily driven by strong advancements in flowering in the 1990s and early 2000s.

CONCLUSIONS

Analysis of long-term trends can mask substantial variation in phenological shifts through time. This variation in the direction and magnitude of phenological shifts has implications for the evolution of flowering time and for interspecific interactions with flowering plants and can provide more detailed insights into the dynamics of phenological responses to climate change.

A refined modelling approach to assess the influence of sampling on palaeobiodiversity curves: new support for declining Cretaceous dinosaur richness

Modelling has been underdeveloped with respect to constructing palaeobiodiversity curves, but it offers an additional tool for removing sampling from their estimation. Here, an alternative to subsampling approaches, which often require large sample sizes, is explored by the extension and refinement of a pre-existing modelling technique that uses a geological proxy for sampling. Application of the model to the three main clades of dinosaurs suggests that much of their diversity fluctuations cannot be explained by sampling alone. Furthermore, there is new support for a long-term decline in their diversity leading up to the Cretaceous-Paleogene (K-Pg) extinction event. At present, use of this method with data that includes either Lagerstätten or 'Pull of the Recent' biases is inappropriate, although partial solutions are offered.

Large-scale modeling - a tool for conquering the complexity of the brain

Is there any hope of achieving a thorough understanding of higher functions such as perception, memory, thought and emotion or is the stunning complexity of the brain a barrier which will limit such efforts for the foreseeable future? In this perspective we discuss methods to handle complexity, approaches to model building, and point to detailed large-scale models as a new contribution to the toolbox of the computational neuroscientist. We elucidate some aspects which distinguishes large-scale models and some of the technological challenges which they entail.

Improving the accuracy of sampling field plots for plant-parasitic nematodes

The validity of nematode data from field experiments depends largely on how well samples represent the nematode population. Data from an intensive sampling of three field plots before and after spring cultivation were used to compare eight simulated sampling schemes. Average deviation from the plot mean ranged from 10% to 34% before cultivation and from 7% to 16% after cultivation. Samples taken from only the plant row erred most before cultivation but were comparable to other schemes after cultivation. Several schemes achieved a 25% deviation or less in 90% of the sample simulations. Sampling a nematode population usually involves subsampling a composite bulk sample, however, and this increases error by an estimable amount. A random sample with 35 cores and four random subsamples estimated mean plot densities within 25% with probabilities ranging from 0.77 to 0.85. The probability of a sample-subsample combination coming within a specified percent error of the true mean can be extended cautiously to any field mean and variance more-or-less independent of species and area using formulae presented herein. The most economical method of increasing sample accuracy was to increase the number of soil cores.