Pattern Detection in Intensive Care Monitoring Time Series with Autoregressive Models: Influence of the Model Order

The detection of patterns in monitoring data of vital signs is of great importance for adequate bedside decision support in critical care. Currently used alarm systems, which are based on fixed thresholds and independency assumptions, are not satisfactory in clinical practice. Time series techniques such as AR‐models consider autocorrelations within the series, which can be used for pattern recognition in the data. For practical applications in intensive care the data analysis has to be automated. An important issue is the suitable choice of the model order which is difficult to accomplish online. In a comparative case‐study we analyzed 34564 univariate time series of hemodynamic variables in critically ill patients by autoregressive models of different orders and compared the results of pattern detection. AR(2)‐models seem to be most suitable for the detection of clinically relevant patterns, thus affirming that treating the data as independent leads to false alarms. Moreover, using AR(2)‐models affords only short estimation periods. These findings for pattern detection in intensive care data are of medical importance as they justify a preselection of a model order, easing further automated statistical online analysis.     

Online sequential extreme studentized deviate tests for anomaly detection in streaming data with varying patterns

In the new era of big data, numerous information and technology systems can store huge amounts of streaming data in real time, for example, in server-access logs on web application servers. The importance of anomaly detection in voluminous quantities of streaming data from such systems is rapidly increasing. One of the biggest challenges in the detection task is to carry out real-time contextual anomaly detection in streaming data with varying patterns that are visually detectable but unsuitable for a parametric model. Most anomaly detection algorithms have weaknesses in dealing with streaming time-series data containing such patterns. In this paper, we propose a novel method for online contextual anomaly detection in streaming time-series data using generalized extreme studentized deviates (GESD) tests. The GESD test is relatively accurate and efficient because it performs statistical hypothesis testing but it is unable to handle streaming time-series data. Thus, focusing on streaming time-series data, we propose an online version of the test capable of detecting outliers under varying patterns. We perform extensive experiments with simulated data, syntactic data, and real online traffic data from Yahoo Webscope, showing a clear advantage of the proposed method, particularly for analyzing streaming data with varying patterns.

     

A Stochastic Regression Model for General Trend Analysis of Longitudinal Continuous Data

A predictive continuous time model is developed for continuous panel data to assess the effect of time‐varying covariates on the general direction of the movement of a continuous response that fluctuates over time. This is accomplished by reparameterizing the infinitesimal mean of an Ornstein–Uhlenbeck processes in terms of its equilibrium mean and a drift parameter, which assesses the rate that the process reverts to its equilibrium mean. The equilibrium mean is modeled as a linear predictor of covariates. This model can be viewed as a continuous time first‐order autoregressive regression model with time‐varying lag effects of covariates and the response, which is more appropriate for unequally spaced panel data than its discrete time analog. Both maximum likelihood and quasi‐likelihood approaches are considered for estimating the model parameters and their performances are compared through simulation studies. The simpler quasi‐likelihood approach is suggested because it yields an estimator that is of high efficiency relative to the maximum likelihood estimator and it yields a variance estimator that is robust to the diffusion assumption of the model. To illustrate the proposed model, an application to diastolic blood pressure data from a follow‐up study on cardiovascular diseases is presented. Missing observations are handled naturally with this model.     

Tests for Monotonic Trend from Case‐Control Data: Cochran‐Armitage‐Mantel Trend Test,Isotonic Regression and Single and Multiple Contrast Tests

Tests for a monotonic trend between an ordered categorical exposure and disease status are routinely carried out from case‐control data using the Mantel‐extension trend test or the asymptotically equivalent Cochran‐Armitage test. In this study, we considered two alternative tests based on isotonic regression, namely an order‐restricted likelihood ratio test and an isotonic modification of the Mantel‐extension test extending the recent proposal by Mancuso, Ahn and Chen (2001) to case‐control data. Furthermore, we considered three tests based on contrasts, namely a single contrast (SC) test based on Schaafsma's coefficients, the Dosemeci and Benichou (DB) test, a multiple contrast (MC) test based on the Helmert, reverse‐Helmert and linear contrasts and we derived their case‐control versions. Using simulations, we compared the statistical properties of these five alternative tests to those of the Mantel‐extension test under various patterns including no relationship, as well as monotonic and non‐monotonic relationships between exposure and disease status. In the case of no relationship, all tests had close to nominal type I error except in situations combining a very unbalanced exposure distribution and small sample size, where the asymptotic versions of the three tests based on contrasts were highly anticonservative. The use of bootstrap instead of asymptotic versions corrected this anticonservatism. For monotonic patterns, all tests had close powers. For non monotonic patterns, the DB‐test showed the most favourable results as it was the least powerful test. The two tests based on isotonic regression were the most powerful tests and the Mantel‐extension test, the SC‐ and MC‐tests had in‐between powers. The six tests were applied to data from a case‐control study investigating the relationship between alcohol consumption and risk of laryngeal cancer in Turkey. In situations with no evidence of a monotonic relationship between exposure and disease status, the three tests based on contrasts did not conclude in favour of a significant trend whereas all the other tests did. (© 2004 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

Mapping and explaining wolf recolonization in France using dynamic occupancy models and opportunistic data

While large carnivores are recovering in Europe, assessing their distributions can help to predict and mitigate conflicts with human activities. Because they are highly mobile, elusive and live at very low density, modeling their distributions presents several challenges due to 1) their imperfect detectability, 2) their dynamic ranges over time and 3) their monitoring at large scales consisting mainly of opportunistic data without a formal measure of the sampling effort. Here, we focused on wolves Canis lupus that have been recolonizing France since the early 1990s. We evaluated the sampling effort a posteriori as the number of observers present per year in a cell based on their location and professional activities. We then assessed wolf range dynamics from 1994 to 2016, while accounting for species imperfect detection and time‐ and space‐varying sampling effort using dynamic site‐occupancy models. Ignoring the effect of sampling effort on species detectability led to underestimating the number of occupied sites by more than 50% on average. Colonization appeared to be negatively influenced by the proportion of a site with an altitude higher than 2500 m and positively influenced by the number of observed occupied sites at short and long‐distances, forest cover, farmland cover and mean altitude. The expansion rate, defined as the number of occupied sites in a given year divided by the number of occupied sites in the previous year, decreased over the first years of the study, then remained stable from 2000 to 2016. Our work shows that opportunistic data can be analyzed with species distribution models that control for imperfect detection, pending a quantification of sampling effort. Our approach has the potential for being used by decision‐makers to target sites where large carnivores are likely to occur and mitigate conflicts.     

Analysis of dose-response effects on gene expression data with comparison of two microarray platforms

MOTIVATION: The problems of analyzing dose effects on gene expression are gaining attention in biomedical research. A specific challenge is to detect genes with expression levels that change according to dose levels in a non-random manner, but nonetheless may be considered as potential biomarkers. METHOD: We are among the first to formally apply a tool that uses an isotonic (monotonic) regression approach to this area of study. We introduce a test statistic to select genes with significant dose-response expression in a monotonic fashion based on a permutation procedure. We then compare the results with those achieved from the application of a likelihood ratio-based test. RESULTS: We apply the isotonic regression approach to a study of gene expression in the RKO colon carcinoma cell line in response to varying dosage levels of the chemotherapeutic agent 5-fluorouracil. A feature of both Affymetrix and printed 75mer oligomer cDNA arrays produced from the same samples provides an opportunity to compare the two microarray platforms. AVAILABILITY: Statistical software S-plus Code to implement the method is available from the authors. CONTACT: kcoombes@mdanderson.org     

Modelling time to population extinction when individual reproduction is autocorrelated

In nature, individual reproductive success is seldom independent from year to year, due to factors such as reproductive costs and individual heterogeneity. However, population projection models that incorporate temporal autocorrelations in individual reproduction can be difficult to parameterise, particularly when data are sparse. We therefore examine whether such models are necessary to avoid biased estimates of stochastic population growth and extinction risk, by comparing output from a matrix population model that incorporates reproductive autocorrelations to output from a standard age‐structured matrix model that does not. We use a range of parameterisations, including a case study using moose data, treating probabilities of switching reproductive class as either fixed or fluctuating. Expected time to extinction from the two models is found to differ by only small amounts (under 10%) for most parameterisations, indicating that explicitly accounting for individual reproductive autocorrelations is in most cases not necessary to avoid bias in extinction estimates.     

Measurement error and estimates of population extinction risk

It is common to estimate the extinction probability for a vulnerable population using methods that are based on the mean and variance of the long‐term population growth rate. The numerical values of these two parameters are estimated from time series of population censuses. However, the proportion of a population that is registered at each census is typically not constant but will vary among years because of stochastic factors such as weather conditions at the time of sampling. Here, we analyse how such sampling errors influence estimates of extinction risk and find sampling errors to produce two opposite effects. Measurement errors lead to an exaggerated overall variance, but also introduce negative autocorrelations in the time series (which means that estimates of annual growth rates tend to alternate in size). If time series data are treated properly these two effects exactly counter balance. We advocate routinely incorporating a measure of among year correlations in estimating population extinction risk.     

1982-2018年中国植被覆盖变化非线性趋势及其格局分析

探究植被覆盖变化是评估陆地生态系统环境变化的重要手段,但现有研究多采用线性趋势来表达植被覆盖的变化情况而忽略了趋势的非线性。本文使用GLASS FVC数据,利用BFAST方法和格局分析,探讨了1982-2018年我国植被覆盖变化的非线性趋势及其分布格局。结果表明：（1）与线性趋势方法的对比发现,BFAST的检测结果揭示了四川盆地、黄土高原等地的植被覆盖显著增加趋势其实存在中断,青海和东北等地植被覆盖经历了由退化到改善的过程而并非简单的线性增加,而青藏高原中东部等地则由原先的改善趋势变为了退化趋势。（2）将非线性趋势结果进行分类,其中单调型增加类型占比最多,达到33.58%,主要分布在内蒙古、陕西及河南等地;单调型减少占比1.82%,主要分布在东南沿海地区;中断型增加占比22.91%,主要分布在四川盆地东部和华北地区;中断型减少占比2.68%,主要分布在青藏高原东南部;由增到减占比4.20%,主要分布在青海等地;由减到增占比14.62%,主要分布在吉林等地。大范围的植被覆盖增加趋势充分反映了我国过去几十年植被的改善,但同时存在的减少趋势表明潜在的植被退化风险仍不可忽视。（3）不同趋势类型发生改变的时间有所差异,总体上1988-1999年间发生的改变较少,而2000-2011年间发生的改变较多,我国21世纪以来实施的大规模生态保护和恢复工程对植被的改善过程有重要影响。（4）分布格局上,植被覆盖改善趋势类型（单调型增加,中断型增加,由减到增）呈现大聚集,小分散的特点,具有复杂的形状;退化趋势类型（单调型减少,中断型减少,由增到减）的面积均较小,分布也相对离散。全国尺度上趋势空间格局呈现一定规律但分布的异质性较大,区域尺度上植被覆盖经受的干扰显著,变化过程实际也是较为复杂的。本研究表明,使用非线性趋势方法和格局分析,可以更准确地评估植被覆盖的时空变化,从而为生态环境相关工作的开展提供科学的参考。     

Statistical analysis of personal radiofrequency electromagnetic field measurements with nondetects

Exposimeters are increasingly applied in bioelectromagnetic research to determine personal radiofrequency electromagnetic field (RF‐EMF) exposure. The main advantages of exposimeter measurements are their convenient handling for study participants and the large amount of personal exposure data, which can be obtained for several RF‐EMF sources. However, the large proportion of measurements below the detection limit is a challenge for data analysis. With the robust ROS (regression on order statistics) method, summary statistics can be calculated by fitting an assumed distribution to the observed data. We used a preliminary sample of 109 weekly exposimeter measurements from the QUALIFEX study to compare summary statistics computed by robust ROS with a naïve approach, where values below the detection limit were replaced by the value of the detection limit. For the total RF‐EMF exposure, differences between the naïve approach and the robust ROS were moderate for the 90th percentile and the arithmetic mean. However, exposure contributions from minor RF‐EMF sources were considerably overestimated with the naïve approach. This results in an underestimation of the exposure range in the population, which may bias the evaluation of potential exposure‐response associations. We conclude from our analyses that summary statistics of exposimeter data calculated by robust ROS are more reliable and more informative than estimates based on a naïve approach. Nevertheless, estimates of source‐specific medians or even lower percentiles depend on the assumed data distribution and should be considered with caution. Bioelectromagnetics. Bioelectromagnetics 29:471–478, 2008. © 2008 Wiley‐Liss, Inc.     

Empirical patterns of environmental variation favor adaptive transgenerational plasticity

Effects of parental environment on offspring traits have been well known for decades. Interest in this transgenerational form of phenotypic plasticity has recently surged due to advances in our understanding of its mechanistic basis. Theoretical research has simultaneously advanced by predicting the environmental conditions that should favor the adaptive evolution of transgenerational plasticity. Yet whether such conditions actually exist in nature remains largely unexplored. Here, using long‐term climate data, we modeled optimal levels of transgenerational plasticity for an organism with a one‐year life cycle at a spatial resolution of 4 km² across the continental United States. Both annual temperature and precipitation levels were often autocorrelated, but the strength and direction of these autocorrelations varied considerably even among nearby sites. When present, such environmental autocorrelations render offspring environments statistically predictable based on the parental environment, a key condition for the adaptive evolution of transgenerational plasticity. Results of our optimality models were consistent with this prediction: High levels of transgenerational plasticity were favored at sites with strong environmental autocorrelations, and little‐to‐no transgenerational plasticity was favored at sites with weak or nonexistent autocorrelations. These results are among the first to show that natural patterns of environmental variation favor the evolution of adaptive transgenerational plasticity. Furthermore, these findings suggest that transgenerational plasticity is likely variable in nature, depending on site‐specific patterns of environmental variation.     

An algorithm for online detection of temporal changes in operator cognitive state using real-time psychophysiological data

We consider the problem of on-the-fly detection of temporal changes in the cognitive state of human subjects due to varying levels of difficulty of performed tasks using real-time EEG and EOG data. We construct the Cognitive State Indicator (CSI) as a function that projects the multidimensional EEG/EOG signals onto the interval [0,1] by maximizing the Kullback–Leibler distance between distributions of the signals, and whose values change continuously with variations in cognitive load. During offline testing (i.e., when evolution in time is disregarded) it was demonstrated that the CSI can serve as a statistically significant discriminator between states of different cognitive loads. In the online setting, a trend detection heuristic (TDH) has been proposed to detect real-time changes in the cognitive state by monitoring trends in the CSI. Our results support the application of the CSI and the TDH in future closed-loop control systems with human supervision.     

One-step method for plasma determination of ibuprofen by chemiluminescence-coupled ultrafiltration and application in a pharmacokinetic study

A simple one‐step method is established for plasma determination of ibuprofen and its pharmacokinetic study. The method involves simple sample pre‐treatment by dilution, rapid separation by ultrafiltration (UF) and online sensitive detection by chemiluminescence (CL) based on significant intensity enhancement of ibuprofen on the weak CL of potassium permanganate and sodium sulphite in an acidic system. The calibration curve for ibuprofen is linear in the range 0.1–50.0 µg/mL in rat plasma. Average recoveries of ibuprofen at 0.80, 12.0 and 40.0 µg/mL amounted to 98.0 ± 4.2%, 101.2 ± 3.6% and 99.3 ± 5.4%, respectively. Standard deviations of intra‐ and inter‐day measurement precision and accuracy are within ±10.0%. The detection limit for ibuprofen is 10.0 µg/L in plasma samples. Pharmacokinetic study of ibuprofen by the validated method shows that the mean plasma drug concentration–time course confirms to a classical two‐compartment open model with first‐order absorption. The proposed method will be an alternative for pre‐clinical pharmacokinetic study of ibuprofen and other non‐steroidal anti‐inflammatory drugs. Copyright © 2011 John Wiley & Sons, Ltd.     

No consistent effects of humans on animal genetic diversity worldwide

Human impacts on genetic diversity are poorly understood yet critical to biodiversity conservation. We used 175 247 COI sequences collected between 1980 and 2016 to assess the global effects of land use and human density on the intraspecific genetic diversity of 17 082 species of birds, fishes, insects and mammals. Human impacts on mtDNA diversity were taxon and scale‐dependent, and were generally weak or non‐significant. Spatial analyses identified weak latitudinal diversity gradients as well as negative effects of human density on insect diversity, and negative effects of intensive land use on fish diversity. The observed effects were predominantly associated with species turnover. Time series analyses found nearly an equal number of positive and negative temporal trends in diversity, resulting in no net monotonic trend in diversity over this time period. Our analyses reveal critical data and theory gaps and call for increased efforts to monitor global genetic diversity.     

Spatio‐temporal modelling of Glossina palpalis gambiensis and Glossina tachinoides apparent densities in fragmented ecosystems of Burkina Faso

Tsetse are the insect vectors of the African trypanosomiases. As with many diseases, transmission of trypanosomiasis varies through space and time. Capturing the variation of both vector and disease has, in the past, been attempted separately in the space and time dimensions, usually using deterministic techniques. Very few efforts have used space‐time covariation and have hence missed any correlations that may exist between variation in these two dimensions. Here we propose two novel approaches to space‐time analysis derived from space‐time geostatistics in a kriging framework. The approaches were developed through analysis of a dataset recording the Apparent Density of Glossina palpalis gambiensis and Glossina tachinoides (Diptera: Glossinidae) in three riparian sites in Burkina Faso over 15 months between 2006 and 2007. This site is fragmented due to human activity in the area. The first approach, Space Time Ordinary Kriging, does not consider the effect of fragmentation. It is used as a benchmark to test the increased explanatory power of the second method, which does account for fragmentation. The second method, Regression Space Time Simple Kriging, is a distinct improvement over the first approach because it allows for a spatial trend in the mean trap catch; this trend is related to, and later predicted from, environmental co‐variates. The results indicate the presence of space and time effects on tsetse distribution, dependent on the size of the habitat fragmentation patches. These effects occur at relatively small geographic scales within a season. Whilst such variation has long been suspected, the new methods presented here are able to quantify this variation precisely, so that seasonal and spatial comparisons can now be made both within and between species.     

Innovation in teaching and learning invertebrate zoology in remote and online classrooms

The COVID‐19 global pandemic caused instructors to pivot to remote and online teaching, an especially challenging task in hands‐on classes such as invertebrate biology. In this special 25th anniversary issue of Invertebrate Biology, the authors present a variety of clever and effective ways to help invertebrate biology instructors adapt to teaching in an online environment. Student‐centered research and learning are essential in all biology classes, and we explore scientific writing, field trips, do‐it‐yourself laboratories, and more. These techniques will be useful for classes of varying sizes and types, from non‐major undergraduates to graduate students, even after the pandemic is over. Innovation for teaching invertebrate biology online may help facilitate more inclusive courses that serve diverse students more equitably. Ideas for how to best move traditionally hands‐on laboratories into online or remote formats are currently also being informally discussed in a collaborative online space for instructors.     

Maximization principles for frequency-dependent selection I: the one-locus two-allele case

In this article we study the one-locus two-allele version of the pairwise-interaction model of frequency-dependent selection in discrete and continuous time. Our main aim is to provide necessary and sufficient conditions for the validity of maximization principles. We provide a systematic approach that covers all possible facets of the dynamical behavior of the model, and we illustrate our results by concrete examples. We show that the mean fitness of the population is nondecreasing if the interaction coefficients are symmetric and positive. Moreover, monotonic convergence to the set of equilibria always occurs, which is not true if we also consider negative interaction coefficients. For asymmetric interaction, we provide necessary conditions when the mean fitness is nondecreasing and sufficient conditions when it is not. Furthermore, in discrete time, we show that limit cycles cannot occur, unless some interaction coefficients are negative.     

Detecting long‐term growth trends using tree rings: a critical evaluation of methods

Tree‐ring analysis is often used to assess long‐term trends in tree growth. A variety of growth‐trend detection methods (GDMs) exist to disentangle age/size trends in growth from long‐term growth changes. However, these detrending methods strongly differ in approach, with possible implications for their output. Here, we critically evaluate the consistency, sensitivity, reliability and accuracy of four most widely used GDMs: conservative detrending (CD) applies mathematical functions to correct for decreasing ring widths with age; basal area correction (BAC) transforms diameter into basal area growth; regional curve standardization (RCS) detrends individual tree‐ring series using average age/size trends; and size class isolation (SCI) calculates growth trends within separate size classes. First, we evaluated whether these GDMs produce consistent results applied to an empirical tree‐ring data set of Melia azedarach, a tropical tree species from Thailand. Three GDMs yielded similar results – a growth decline over time – but the widely used CD method did not detect any change. Second, we assessed the sensitivity (probability of correct growth‐trend detection), reliability (100% minus probability of detecting false trends) and accuracy (whether the strength of imposed trends is correctly detected) of these GDMs, by applying them to simulated growth trajectories with different imposed trends: no trend, strong trends (?6% and +6% change per decade) and weak trends (?2%, +2%). All methods except CD, showed high sensitivity, reliability and accuracy to detect strong imposed trends. However, these were considerably lower in the weak or no‐trend scenarios. BAC showed good sensitivity and accuracy, but low reliability, indicating uncertainty of trend detection using this method. Our study reveals that the choice of GDM influences results of growth‐trend studies. We recommend applying multiple methods when analysing trends and encourage performing sensitivity and reliability analysis. Finally, we recommend SCI and RCS, as these methods showed highest reliability to detect long‐term growth trends.     

Non‐invasive online detection of microbial lysine formation in stirred tank bioreactors by using calorespirometry

Non‐invasive methods for online monitoring of biotechnological processes without compromising the integrity of the reactor system are very important to generate continuous data. Even though calorimetry has been used in conventional biochemical analysis for decades, it has not yet been specifically applied for online detection of product formation at technical scale. Thus, this article demonstrates a calorespirometric method for online detection of microbial lysine formation in stirred tank bioreactors. The respective heat generation of two bacterial strains, Corynebacterium glutamicum ATCC 13032 (wild‐type) and C. glutamicum DM1730 (lysine producer), was compared with the O₂‐consumption in order to determine whether lysine was formed. As validation of the proposed calorespirometric method, the online results agreed well with the offline measured data. This study has proven that calorespirometry is a viable non‐invasive technique to detect product formation at any time point. Biotechnol. Bioeng. 2013; 110: 1386–1395. © 2012 Wiley Periodicals, Inc.     

Mixture Models for Distance Sampling Detection Functions

We present a new class of models for the detection function in distance sampling surveys of wildlife populations, based on finite mixtures of simple parametric key functions such as the half-normal. The models share many of the features of the widely-used “key function plus series adjustment” (K+A) formulation: they are flexible, produce plausible shapes with a small number of parameters, allow incorporation of covariates in addition to distance and can be fitted using maximum likelihood. One important advantage over the K+A approach is that the mixtures are automatically monotonic non-increasing and non-negative, so constrained optimization is not required to ensure distance sampling assumptions are honoured. We compare the mixture formulation to the K+A approach using simulations to evaluate its applicability in a wide set of challenging situations. We also re-analyze four previously problematic real-world case studies. We find mixtures outperform K+A methods in many cases, particularly spiked line transect data (i.e., where detectability drops rapidly at small distances) and larger sample sizes. We recommend that current standard model selection methods for distance sampling detection functions are extended to include mixture models in the candidate set.