Bias Covariance Research Articles

Atmospheric bias teleconnections in boreal winter associated with systematic sea surface temperature errors in the tropical Indian Ocean

Abstract. Coupled climate models suffer from significant sea surface temperature (SST) biases in the tropical Indian Ocean (TIO), leading to errors in global climate predictions. In this study, we investigate the local and remote effects of the TIO SST bias on the simulated atmospheric circulation and spatio-temporal variability – bias teleconnections. A set of century-long simulations forced by idealized SST perturbations, which resemble various (monopolar or dipolar, positive or negative) TIO SST biases in coupled climate models, are conducted with an intermediate-complexity atmospheric model. Bias teleconnections with a focus on boreal wintertime are analysed using the normal-mode function (NMF) decomposition, which can differentiate between balanced and unbalanced flows across spatial scales. The results show that the atmospheric circulation biases caused by the TIO SST bias have the Gill–Matsuno-type pattern in the tropics and Rossby-wave-train structure in the extratropics, similar to the steady-state response to tropical heating perturbations. The teleconnections between the tropical and extratropical biases are set up by Rossby wave activity flux emanating from the subtropics. Over 90 % of the bias variance (i.e. the square of the bias amplitude) is contained in zonal wavenumbers k≤5. The northward shift of the SST bias away from the Equator weakens the amplitude but does not change the spatial structure of the atmospheric response. Besides, the positive SST bias produces stronger bias teleconnections than the negative one of the same size and magnitude. In the NMF framework, the change in the spatial variance of the time-mean state (i.e. energy) is equal to the sum of the bias variance and the covariance between the circulation bias and the reference state (i.e. bias covariance). Due to the TIO SST biases, the global unbalanced zonal-mean (k=0) flow energy decreases, whereas its balanced counterpart increases. These changes primarily arise from the strong bias covariance. For k>0, both the global unbalanced and the tropical balanced energies increase in the case of a monopolar SST bias and decrease in the case of a dipolar SST bias. The increase appears mainly as the bias variance, whereas the decrease is associated with a strong negative bias covariance at k=1 and 2. In contrast, the extratropical balanced wave energy decreases (increases) when the TIO SST bias is positive (negative), which is mainly associated with the bias covariance at k=1. The change in the interannual variance (IAV) is contingent upon the sign of the TIO SST bias. A positive bias reduces, whereas a negative one increases, the IAV in both balanced and unbalanced flows. Geographically, large IAV changes are observed in the tropical Indo-West Pacific region, Australia, South and Northeast Asia, the Pacific-North America region, and Europe, where the background IAVs are strong.

Deep Texture-Aware Features for Camouflaged Object Detection

Camouflaged object detection is a challenging task that aims to identify objects having similar texture to the surroundings. This paper presents to amplify the subtle texture difference between camouflaged objects and the background for camouflaged object detection by formulating multiple texture-aware refinement modules to learn the texture-aware features in a deep convolutional neural network. The texture-aware refinement module computes the biased co-variance matrices of feature responses to extract the texture information, adopts an affinity loss to learn a set of parameter maps that help to separate the texture between camouflaged objects and the background, and leverages a boundary-consistency loss to explore the structures of object details. We evaluate our network on the benchmark datasets for camouflaged object detection both qualitatively and quantitatively. Experimental results show that our approach outperforms various state-of-the-art methods by a large margin.

Bayesian inverse modeling of large-scale spatial fields on iteratively corrected principal components

Inverse modeling of large-scale spatially variable parameters fields at fine resolution can be reduced to estimating the projections on dominant principal components of the underlying parameter fields based on principal component analysis of the spatial covariance. For unknown or biased prior structural parameters of spatial covariance models, an iterative procedure consisting of two successive steps is usually implemented, i.e., estimation of spatial covariance followed by estimation of the spatially variable parameter fields conditional on the spatial covariance and observations. In this study, we develop an iterative, computationally efficient method to update dominant principal components for nonlinear inverse problems of large-scale spatial fields that adaptively corrects the bias from the initially defined prior spatial covariance. Our algorithm involves two-layer iterations: the inner iteration is to obtain the best estimates of projections on given retained principal components, and the outer iteration implements an efficient rank-one updating to correct the retained principal components using the posterior covariance associated with the best estimates of the projections. Numerical experiments show that inversion results can be significantly improved for large-scale inverse problems with biased structural parameters for spatial covariance. The experiment results show that the iterative correction is essentially to match the distribution patterns of the spatially correlated parameter field with its most dominant principal components. We also investigate the performance of the developed method under different biased covariance model initialization, including model type bias, variance bias and correlation length bias. The correction cannot fundamentally change the smoothness defined by the covariance model type, but can still describe major distribution patterns including anisotropy. Biased variance can be corrected and yields similar best estimates and variance maps, and biased correlation length can be corrected within an applicable range.

Robust partly strong tracking consider SDRE filter for direct INS/GNSS integration with biases

In order to counteract the adverse effects of biases in direct inertial navigation system/global navigation satellite system integration, a novel robust partly strong tracking consider state-dependent Riccati equation filter (PSTCSDREF) algorithm is proposed. A nonlinear ‘consider’ approach is utilized to incorporate bias statistics into a state estimation error covariance of the state-dependent Riccati equation filter (SDREF), and a new consider SDREF (CSDREF) is proposed. Next, the prediction covariance of the state is partly multiplied by a strong tracking factor, which does not include bias covariance, so as to mitigate the adverse effects of model uncertainties on navigation performance. Numerical simulation demonstrates that the proposed PSTCSDREF demonstrates better navigational accuracy as compared with the extended Kalman filter, SDREF, and CSDREF.

Psychometric Modelling of Longitudinal Genetically Informative Twin Data.

The often-used A(C)E model that decomposes phenotypic variance into parts due to additive genetic and environmental influences can be extended to a longitudinal model when the trait has been assessed at multiple occasions. This enables inference about the nature (e.g., genetic or environmental) of the covariance among the different measurement points. In the case that the measurement of the phenotype relies on self-report data (e.g., questionnaire data), often, aggregated scores (e.g., sum–scores) are used as a proxy for the phenotype. However, earlier research based on the univariate ACE model that concerns a single measurement occasion has shown that this can lead to an underestimation of heritability and that instead, one should prefer to model the raw item data by integrating an explicit measurement model into the analysis. This has, however, not been translated to the more complex longitudinal case. In this paper, we first present a latent state twin A(C)E model that combines the genetic twin model with an item response theory (IRT) model as well as its specification in a Bayesian framework. Two simulation studies were conducted to investigate 1) how large the bias is when sum–scores are used in the longitudinal A(C)E model and 2) if using the latent twin model can overcome the potential bias. Results of the first simulation study (e.g., AE model) demonstrated that using a sum–score approach leads to underestimated heritability estimates and biased covariance estimates. Surprisingly, the IRT approach also lead to bias, but to a much lesser degree. The amount of bias increased in the second simulation study (e.g., ACE model) under both frameworks, with the IRT approach still being the less biased approach. Since the bias was less severe under the IRT approach than under the sum–score approach and due to other advantages of latent variable modelling, we still advise researcher to adopt the IRT approach. We further illustrate differences between the traditional sum–score approach and the latent state twin A(C)E model by analyzing data of a two-wave twin study, consisting of the answers of 8,016 twins on a scale developed to measure social attitudes related to conservatism.

An improved real-time adaptive Kalman filter with recursive noise covariance updating rules

Kalman filter (KF) is used extensively for state estimation. Among its requirements are the process and observation noise covariances which are unknown or partially known in real life applications. Biased initializations of the covariances result in performance degradation of KF or divergence. Therefore, an extensive research is carried on to improve its performance however, relying on a moving window, heavy computations, and the availability of the exact model are the fundamental problems in most of the proposed techniques. In this paper, we are using the idea of the recursive estimation of KF to propose two recursive updating rules for the process and observation covariances respectively designed based on the covariance matching principles. Each rule is a tuned scaled version of the previous covariance in addition to a tuned correction term derived based on the most recent available data. The proposed adaptive Kalman filter AKF avoided the aforementioned problems and proved itself to have an improved performance over the conventional KF. The results show that the AKF estimates are more accurate, have less noise and more stable against biased covariance initializations.

A Simple Attitude Unscented Kalman Filter: Theory and Evaluation in a Magnetometer-Only Spacecraft Scenario

A quaternion-based attitude unscented Kalman filter is formulated with quaternion errors parameterized by small angle approximations and is applied to a filter with a state vector consisting of the attitude quaternion and the gyro bias vector. The filter is evaluated using extensive Monte Carlo data in a simulated lost-in-space scenario of a low-Earth orbiting spacecraft processing only three-axis magnetometer and gyro measurements. The filter is found to be robust, accurate, and rapidly convergent in this scenario for small true gyro biases and small initial uncertainties in their values, often converging in only one half of an orbit period to an attitude accuracy of 0.1 degrees. The filter convergence is found to depend significantly on the value of the true gyro biases as well as the initial gyro bias covariances. Monte Carlo results also indicate that this unscented Kalman filter is significantly less robust than an extended Kalman filter with the same attitude approach, but performs slightly better than another unscented Kalman filter with a generalized Rodrigues parameter approach to quaternion errors.

Robust forecasting of dynamic conditional correlation GARCH models

Large one-off events cause large changes in prices, but may not affect the volatility and correlation dynamics as much as smaller events. In such cases, standard volatility models may deliver biased covariance forecasts. We propose a multivariate volatility forecasting model that is accurate in the presence of large one-off events. The model is an extension of the dynamic conditional correlation (DCC) model. In our empirical application to forecasting the covariance matrix of the daily EUR/USD and Yen/USD return series, we find that our method produces more precise out-of-sample covariance forecasts than the DCC model. Furthermore, when used in portfolio allocation, it leads to portfolios with similar return characteristics but lower turnovers, and hence higher profits.

Forecasting Multivariate Volatility using the VARFIMA Model on Realized Covariance Cholesky Factors

Summary This paper analyzes the forecast accuracy of the multivariate realized volatility model introduced by Chiriac and Voev (2010), subject to different degrees of model parametrization and economic evaluation criteria. Bymodelling the Cholesky factors of the covariance matrices, the model generates positive definite, but biased covariance forecasts. In this paper, we provide empirical evidence that parsimonious versions of the model generate the best covariance forecasts in the absence of bias correction. Moreover, we show by means of stochastic dominance tests that any risk averse investor, regardless of the type of utility function or return distribution, would be better-off from using this model than from using some standard approaches.

Robust Forecasting of Dynamic Conditional Correlation GARCH Models

Large once-off events cause large changes in prices but may not affect volatility and correlation dynamics as much as smaller events. Standard volatility models may deliver biased covariance forecasts in this case. We propose a multivariate volatility forecasting model that is accurate in the presence of large once-off events. The model is an extension of the dynamic conditional correlation model (DCC) model. Compared to the DCC model, our method produces more precise out-of-sample covariance forecasts and, when used in portfolio allocation, it leads to portfolios with similar return characteristics but lower turnover and hence higher profits.

Correcting for Bias in Mahalanobis and Log-Likelihood Estimates

Mahalanobis and log-likelihood estimates are used extensively in tracking systems, but are affected by residual bias and by uncertainty in covariance matrices. This paper provides a formal framework that, to some extent, justifies covariance inflation techniques and allows existing fudge factor thresholds to be interpreted in terms of residual bias covariance matrices and covariance matrix uncertainty.

Ensemble Kalman filter assimilation of temperature and altimeter data with bias correction and application to seasonal prediction

Abstract. To compensate for a poorly known geoid, satellite altimeter data is usually analyzed in terms of anomalies from the time mean record. When such anomalies are assimilated into an ocean model, the bias between the climatologies of the model and data is problematic. An ensemble Kalman filter (EnKF) is modified to account for the presence of a forecast-model bias and applied to the assimilation of TOPEX/Poseidon (T/P) altimeter data. The online bias correction (OBC) algorithm uses the same ensemble of model state vectors to estimate biased-error and unbiased-error covariance matrices. Covariance localization is used but the bias covariances have different localization scales from the unbiased-error covariances, thereby accounting for the fact that the bias in a global ocean model could have much larger spatial scales than the random error.The method is applied to a 27-layer version of the Poseidon global ocean general circulation model with about 30-million state variables. Experiments in which T/P altimeter anomalies are assimilated show that the OBC reduces the RMS observation minus forecast difference for sea-surface height (SSH) over a similar EnKF run in which OBC is not used. Independent in situ temperature observations show that the temperature field is also improved. When the T/P data and in situ temperature data are assimilated in the same run and the configuration of the ensemble at the end of the run is used to initialize the ocean component of the GMAO coupled forecast model, seasonal SSH hindcasts made with the coupled model are generally better than those initialized with optimal interpolation of temperature observations without altimeter data. The analysis of the corresponding sea-surface temperature hindcasts is not as conclusive.

Double-quantum biased covariance spectroscopy: Application to the 2D INADEQUATE experiment

A novel processing scheme is presented that converts a two-dimensional double-quantum NMR spectrum into a single-quantum correlation spectrum. The covariance-like spectrum is computed from the 2D Fourier transform spectrum by emphasizing contributions that fulfill the double-quantum condition resulting in a symmetric spectrum that is easier to analyze. The method is demonstrated for the 2D INADEQUATE experiment.

Short Data Record Adaptive Filtering: The Auxiliary-Vector Algorithm

Karystinos, G. N., Qian, H., Medley, M. J., and Batalama, S. N., Short Data Record Adaptive Filtering: The Auxiliary-Vector Algorithm, Digital Signal Processing 12 (2002) 193–222 Based on statistical conditional optimization criteria, we developed an iterative algorithm that starts from the matched filter (or constraint vector) and generates a sequence of filters that converges to the minimum variance distortionless response (MVDR) solution for any positive definite input autocorrelation matrix. Computationally, the algorithm is a simple recursive procedure that avoids explicit matrix inversion, decomposition, or diagonalization operations. When the input autocorrelation matrix is replaced by a conventional sample-average (positive definite) estimate, the algorithm effectively generates a sequence of MVDR filter estimators: The bias converges rapidly to zero and the covariance trace rises slowly and asymptotically to the covariance trace of the familiar sample matrix inversion (SMI) estimator. For short data records, the early, nonasymptotic, elements of the generated sequence of estimators offer favorable bias–covariance balance and are seen to outperform in mean-square estimation error constraint-LMS, RLS-type, and orthogonal multistage decomposition estimates (also called nested Wiener filters) as well as plain and diagonally loaded SMI estimates. The problem of selecting the most successful (in some appropriate sense) filter estimator in the sequence for a given data record is addressed and two data-driven selection criteria are proposed. The first criterion minimizes the cross-validated sample average variance of the filter estimator output. The second criterion maximizes the estimated J-divergence of the filter estimator output conditional distributions. Illustrative interference suppression examples drawn from the communications literature are followed throughout this presentation.

An iterative algorithm for the computation of the MVDR filter

Statistical conditional optimization criteria lead to the development of an iterative algorithm that starts from the matched filter (or constraint vector) and generates a sequence of filters that converges to the minimum-variance-distortionless-response (MVDR) solution for any positive definite input autocorrelation matrix. Computationally, the algorithm is a simple, noninvasive, recursive procedure that avoids any form of explicit autocorrelation matrix inversion, decomposition, or diagonalization. Theoretical analysis reveals basic properties of the algorithm and establishes formal convergence. When the input autocorrelation matrix is replaced by a conventional sample-average (positive definite) estimate, the algorithm effectively generates a sequence of MVDR filter estimators; the bias converges rapidly to zero and the covariance trace rises slowly and asymptotically to the covariance trace of the familiar sample-matrix-inversion (SMI) estimator. In fact, formal convergence of the estimator sequence to the SMI estimate is established. However, for short data records, it is the early, nonasymptotic elements of the generated sequence of estimators that offer favorable bias covariance balance and are seen to outperform in mean-square estimation error, constraint-LMS, RLS-type, orthogonal multistage decomposition, as well as plain and diagonally loaded SMI estimates. An illustrative interference suppression example is followed throughout this presentation.

Rater bias in psychological research: When is it a problem and what can we do about it?

Rater bias is a substantial source of error in psychological research. Bias distorts observed effect sizes beyond the expected level of attenuation due to intrarater error, and the impact of bias is not accurately estimated using conventional methods of correction for attenuation. Using a model based on multivariate generalizability theory, this article illustrates how bias affects research results. The model identifies 4 types of bias that may affect findings in research using observer ratings, including the biases traditionally termed leniency and halo errors. The impact of bias depends on which of 4 classes of rating design is used, and formulas are derived for correcting observed effect sizes for attenuation (due to bias variance) and inflation (due to bias covariance) in each of these classes. The rater bias model suggests procedures for researchers seeking to minimize adverse impact of bias on study findings.