Inference Error Research Articles

On critical thinking and content knowledge: A critique of the assumptions of cognitive load theory

How critical thinking is understood has a logical impact on pedagogy and curriculum design in a critical thinking education. If critical thinking is seen as a complex mix of knowledge, skills and dispositions that can be articulated and made an object of study in themselves, then there is scope for intentional and explicit targeting of such things. If, however, critical thinking is seen as a general skill more or less impermeable to pedagogical influence and best realised through the development of content knowledge, then an explicit focus on developing thinking skills is misplaced and potentially distracting. The latter view is that of cognitive load theory (CLT), a theoretical framing of thinking and learning that has gained traction in education including with practitioners. This paper will explore some of the assumptions of CLT and show that they are either (1) in error, (2) vaguely or confusingly expressed and/or (3) the conclusions drawn within the theory are often errors of inference. Moreover, I point out that important educational research that weakens the claims of CLT is often not engaged with or addressed.

Optimization of network sensor location for full link flow observability considering sensor measurement error

Full link flow observability problem is to identify the minimum set of links to be installed with traffic sensors in a vehicular traffic network, so that unobserved link flows can be fully inferred from the observed link flows based on the correlation between links. Inevitably, the observed link flows are subject to measurement errors, which will accumulate and propagate in the inference of unobserved link flows. This study aims to identify the minimum observed link set for the installation of counting sensors to reach full link flow observability, while minimizing the effect of sensor measurement errors on the inference of unobserved link flows. We propose a network sensor location model considering the propagation of measurement errors in the link flow inference process. Mathematically, we formulate the problem as min–max and min-sum integer linear programs. The objective function attempts directly to minimize the accumulated inference errors caused by the propagation of sensor measurement errors. When the sensors have the uniform measurement error, it is equivalent to minimizing the maximum or cumulative number of observed links required for the link flow inference of each or all unobserved links. The genetic algorithm (GA) is adopted to solve the proposed optimization model, and the results commend the optimal sensor location scheme with the minimum inference error for full link flow observability. The approach is useful for deploying long-term planning and link-based applications in traffic networks.

Design and Application of Model Based on Bayesian Network

At present, Bayesian networks lack consistent algorithms that support structure establishment, parameter learning, and knowledge reasoning, making it impossible to connect knowledge establishment and application processes. In view of this situation, by designing a genetic algorithm coding method suitable for Bayesian network learning, crossover and mutation operators with adjustment strategies, the fitness function for reasoning error feedback can be carried out. Experimental results show that the new algorithm not only simultaneously optimizes the network structure and parameters, but also can adaptively learn and correct inference errors, and has a more satisfactory knowledge inference accuracy rate.

Regression-based causal inference with factorial experiments: estimands, model specifications and design-based properties

Summary Factorial designs are widely used because of their ability to accommodate multiple factors simultaneously. Factor-based regression with main effects and some interactions is the dominant strategy for downstream analysis, delivering point estimators and standard errors simultaneously via one least-squares fit. Justification of these convenient estimators from the design-based perspective requires quantifying their sampling properties under the assignment mechanism while conditioning on the potential outcomes. To this end, we derive the sampling properties of the regression estimators under a wide range of specifications, and establish the appropriateness of the corresponding robust standard errors for Wald-type inference. The results help to clarify the causal interpretation of the coefficients in these factor-based regressions, and motivate the definition of general factorial effects to unify the definitions of factorial effects in various fields. We also quantify the bias-variance trade-off between the saturated and unsaturated regressions from the design-based perspective.

Irrationality of managers’ behavior in corporate financial decision making

Purpose of the study: The subject of consideration is the behavioral aspects of corporate finance. The consideration is devoted to the basic inclinations of psychological nature characteristic of managers. The main purpose of the article is to characterize the irrational behavior of managers in the process of financial decision-making in the enterprise.
 Methodology: The paper was prepared with the use the critical literature review method mainly in the field of behavioral corporate finance.
 Main findings: The discussion shows that behavioral corporate finance focuses mainly on cognitive and motivational-emotional processes in managers, which may occur in various decision areas within which managers make choices. There are three groups of psychological phenomena and inclinations that are characteristic of managers who manage business entities, namely predispositions to systematic inference errors, heuristics, and the presentation effect.
 Application of the study: The presented article refers to the irrational behavior of managers in the process of making financial decisions in the company. It implies reflections in such scientific fields as, among others, economics and finance, management and psychology. The use of the tools of psychology allows analyzing the problems of financial decision-making of managers in the enterprise, noticing in them some deviations from rationality that can affect the efficiency of the enterprise. The content of the article can be useful for managers making financial decisions in an enterprise.
 Originality/Novelty of the study: Behavioral finance is a young discipline of finance, the scientific output of which in Poland is still small. Behavioral aspects are just beginning to gain importance in the decision-making process, especially the financial one. The tendencies of managers to irrationality in the decision-making process presented in this article allow us to better understand the errors, psychological factors that may cause wrong decisions, which in turn may translate into poorer financial condition of the whole enterprise. The article can inspire further research and inquiry in the field of behavioral finance and contribute to other interesting scientific studies.

Deep learning for automated left ventricular outflow tract diameter measurements in 2D echocardiography

Abstract Background Left ventricular outflow tract diameter (LVOTd) is routinely measured to calculate stroke volume and estimate aortic valve area by the continuity equation. Despite LVOTd being regularly measured clinically, significant inter- and intraobserver variability is evident. This variability is highly impactful on both aortic stenosis evaluation and cardiac output calculation due to the squaring of the LVOT radius. Purpose We aimed to investigate if LVOTd measurements from clinical echocardiographic examinations could be used in a deep learning (DL) model to automatically perform LVOTd measurements with equivalent accuracy and improved consistency compared to current practice. Methods Data was collected from clinical echocardiographic examinations performed on 656 consecutive patients admitted to the cardiac catheterization laboratory at a university hospital in January – December 2018. Parasternal views with cardiologist annotated LVOTd coordinates were assessed for 1314 echocardiographic still images. The quality of the still image and annotated LVOT ground truth were individually graded as high, medium and low by experienced cardiologists to establish a rigorous training basis. Spatial geometry data was preserved for each still image in order to distinguish between different degrees of image zoom. Data was randomly split into training, validation and testing sets (68%, 17%, 15%). A fully convolutional network based on the Resnet50 architecture was used with a custom loss function with heatmap regression. Image augmentations were added to extend the dataset. Results When including echocardiographic images of any quality (n=1314) in the model training and inference, the median absolute difference between cardiologist LVOTd and DL LVOTd was 0.97 mm (95% Confidence interval (CI) 0.79–1.14). Using only high and medium quality still images and ground truth (n=869) in the training and inference, median absolute difference decreased to 0.81 mm (95% CI 0.60–0.96). Adding image augmentations to this dataset further improved the model, resulting in a median LVOTd absolute difference of 0.66 mm (95% CI 0.51–0.78). The LVOTd error in inference increased with decreasing image quality, as shown in Figure 1, with two predictions (0.9%) failing completely. Conclusion Deep learning models are capable of measuring LVOTd with comparable accuracy to cardiologists when trained on clinical data. Data quality affects both training and inference. Even with a slightly lower accuracy when used on lower quality echocardiographic images, DL-assisted LVOT measurement has a clear potential to increase repeatability and consistency of LVOTd measurements. Funding Acknowledgement Type of funding sources: Public Institution(s). Main funding source(s): Research Council of Norway (Norges forskningsråd) Figure 1

Identification of diseases and physiological disorders in potato via multispectral drone imagery using machine learning tools

The rapid and precise detection of diseases and plant disorders is the basis for the adequate and timely design of management strategies. Currently, there are several non-destructive alternatives that allow early detection, highlighting the use of spectral cameras attached to unmanned aerial vehicles (UAVs). The objective of this research was to evaluate the use of multispectral cameras on UAVs to discriminate vascular wilt caused by Verticillium spp., (VW), waterlogging stress (WL), and an unknown alteration (UA) in commercial potato (Solanum tuberosum) variety “Diacol Capiro” crops. Plots were monitored during the crop cycle, performing the visual characterization of the diseases and disorders present. Five spectral band images were acquired using a MicaSense RedEdge spectral camera attached to a Map-T680 hexacopter drone to extract the bands and calculate the vegetation indices that were calibrated and evaluated to determine their ability to discriminate between diseased and healthy plants based on a generalized linear model (GLM) and Kappa index. Additionally, the supervised random forest classification method was implemented, optimized, and evaluated using the accuracy, area under receiver operating characteristic curve (ROC-AUC), kappa index, and inference error based on k-fold cross-validation. After algorithms optimization our results show a classifier accuracy, kappa and ROC-AUC values to VW, WL and UA between 73.5–82.5%, 0.56–0.71, 0.97–0.98, and 35 37.5–51.9%, 0.07–0.06, and 0.88–0.94 for plots 1 and 2, respectively. This study reports an approach to the use of multispectral cameras attached to UAVs as a tool with potential for the detection of diseases and physiological disorders in commercial potato crops.

Extracting brain disease-related connectome subgraphs by adaptive dense subgraph discovery.

Group-level brain connectome analysis has attracted increasing interest in neuropsychiatric research with the goal of identifying connectomic subnetworks (subgraphs) that are systematically associated with brain disorders. However, extracting disease-related subnetworks from the whole brain connectome has been challenging, because no prior knowledge is available regarding the sizes and locations of the subnetworks. In addition, neuroimaging data are often mixed with substantial noise that can further obscure informative subnetwork detection. We propose a likelihood-based adaptive dense subgraph discovery (ADSD) model to extract disease-related subgraphs from the group-level whole brain connectome data. Our method is robust to both false positive and false negative errors of edge-wise inference and thus can lead to a more accurate discovery of latent disease-related connectomic subnetworks. We develop computationally efficient algorithms to implement the novel ADSD objective function and derive theoretical results to guarantee the convergence properties. We apply the proposed approach to a brain fMRI study for schizophrenia research and identify well-organized and biologically meaningful subnetworks that exhibit schizophrenia-related salience network centered connectivity abnormality. Analysis of synthetic data also demonstrates the superior performance of the ADSD method for latent subnetwork detection in comparison with existing methods in varioussettings.

On the evolution of a functional approach to memory.

In an analysis of memory systems, Sherry and Schacter (Psychological Review, 94, 439-454, 1987) emphasized the importance of functional and evolutionary considerations for characterizing mechanisms of memory. The present article considers four different yet closely related topics from more recent research in which similar considerations have played a prominent role in shaping both experiment and theory: the seven sins of memory, mechanisms underlying memory misattribution errors, the role of memory in imagining future experiences, and the relation between associative inference and memory errors. These lines of research illustrate the usefulness of attempting to integrate functional and mechanistic considerations, in line with the general approach articulated by Sherry and Schacter.

Noisy neighbors and reticent residents: Distinguishing resident from non-resident individuals to improve passive acoustic monitoring

Passive acoustic monitoring (PAM) is an increasingly common method for studying populations of vocally active species. However, the detection of individuals not resident to a site may obfuscate inferences about occurrence and population change. Here, we provide a framework for distinguishing resident from non-resident individuals to estimate territory—rather than site—occupancy in PAM programs by leveraging datasets on vocal behavior and acoustic detections for spotted owls (Strix occidentalis) in the Sierra Nevada, California, USA. Based on acoustic/GPS tags, the extent over which individuals typically vocalized (the “vocal home range”) was small relative to space use, such that the likelihood of double counting resident territory-holders across multiple survey sites was low. However, comparing passive acoustic detections to known resident owl locations revealed that detections occurred at PAM survey sites known to be unoccupied, possibly because of the presence of non-territorial individuals. Strict thresholds that required acoustic detections over multiple survey periods successfully removed all detections of non-resident individuals but increased the probability of not detecting resident individuals. Conversely, relatively liberal thresholds that minimized the probability of missing resident individuals increased the probability of detecting non-residents. Thus, a tradeoff exists between error types, with optimal threshold criteria dependent on conservation objectives. Our study highlights the importance of examining patterns in individual vocal behavior and acoustic detections to minimize inferential errors in PAM. It also provides a generalizable framework that can be tailored according to specific conservation objectives to strengthen inferences from PAM as it becomes standard practice in conservation science.

Genomic Association vs. Serological Determination of ABO Blood Types in a Chinese Cohort, with Application in Mendelian Randomization.

ABO blood system is an inborn trait determined by the ABO gene. The genetic-phenotypic mechanism underneath the four mutually exclusive and collectively exhaustive types of O, A, B and AB could theoretically be elucidated. However, genetic polymorphisms in the human populations render the link elusive, and importantly, past studies using genetically determined rather than biochemically determined ABO types were not and could not be evaluated for the inference errors. Upon both blood-typing and genotyping a cohort of 1008 people of the Han Chinese population, we conducted a genome-wide association study in parallel with both binomial and multinomial log-linear models. Significant genetic variants are all mapped to the ABO gene, and are quantitatively evaluated for binary and multi-class classification performances. Three single nucleotide polymorphisms of rs8176719, rs635634 and rs7030248 would together be sufficient to establish a multinomial predictive model that achieves high accuracy (0.98) and F1 scores (micro 0.99 and macro 0.97). Using the set of identified ABO-associated genetic variants as instrumental variables, we demonstrate the application in causal analysis by Mendelian randomization (MR) studies on blood pressures (one-sample MR) and severe COVID-19 with respiratory failure (two-sample MR).

RISC-V Virtual Platform-Based Convolutional Neural Network Accelerator Implemented in SystemC

The optimization for hardware processor and system for performing deep learning operations such as Convolutional Neural Networks (CNN) in resource limited embedded devices are recent active research area. In order to perform an optimized deep neural network model using the limited computational unit and memory of an embedded device, it is necessary to quickly apply various configurations of hardware modules to various deep neural network models and find the optimal combination. The Electronic System Level (ESL) Simulator based on SystemC is very useful for rapid hardware modeling and verification. In this paper, we designed and implemented a Deep Learning Accelerator (DLA) that performs Deep Neural Network (DNN) operation based on the RISC-V Virtual Platform implemented in SystemC in order to enable rapid and diverse analysis of deep learning operations in an embedded device based on the RISC-V processor, which is a recently emerging embedded processor. The developed RISC-V based DLA prototype can analyze the hardware requirements according to the CNN data set through the configuration of the CNN DLA architecture, and it is possible to run RISC-V compiled software on the platform, can perform a real neural network model like Darknet. We performed the Darknet CNN model on the developed DLA prototype, and confirmed that computational overhead and inference errors can be analyzed with the DLA prototype developed by analyzing the DLA architecture for various data sets.

Offloading Optimization in Edge Computing for Deep-Learning-Enabled Target Tracking by Internet of UAVs

The empowering unmanned aerial vehicles (UAVs) have been extensively used in providing intelligence such as target tracking. In our field experiments, a pretrained convolutional neural network (CNN) is deployed at UAV to identify a target (a vehicle) from the captured video frames and enable the UAV to keep tracking. However, this kind of visual target tracking demands a lot of computational resources due to the desired high inference accuracy and stringent delay requirement. This motivates us to consider offloading this type of deep learning (DL) tasks to a mobile-edge computing (MEC) server due to the limited computational resource and energy budget of the UAV and further improve the inference accuracy. Specifically, we propose a novel hierarchical DL tasks distribution framework, where the UAV is embedded with lower layers of the pretrained CNN model while the MEC server (MES) with rich computing resources will handle the higher layers of the CNN model. An optimization problem is formulated to minimize the weighted-sum cost, including the tracking delay and energy consumption introduced by communication and computing of UAVs while taking into account the quality of data (e.g., video frames) input to the DL model and the inference errors. Analytical results are obtained and insights are provided to understand the tradeoff between the weighted-sum cost and inference error rate in the proposed framework. Numerical results demonstrate the effectiveness of the proposed offloading framework.

Sensitivity of implicit evaluations to accurate and erroneous propositional inferences

Explicit (directly measured) evaluations are widely assumed to be sensitive to logical structure. However, whether implicit (indirectly measured) evaluations are uniquely sensitive to co-occurrence information or can also reflect logical structure has been a matter of theoretical debate. To test these competing ideas, participants (N = 3928) completed a learning phase consisting of a series of two-step trials. In step 1, one or more conditional statements (A → B) containing novel targets co-occurring with valenced adjectives (e.g., “if you see a blue square, Ibbonif is sincere”) were presented. In step 2, a disambiguating stimulus, e.g., blue square (A) or gray blob (¬A) was revealed. Co-occurrence information, disambiguating stimuli, or both were varied between conditions to enable investigating the unique and joint effects of each. Across studies, the combination of conditional statements and disambiguating stimuli licensed different normatively accurate inferences. In Study 1, participants were prompted to use modus ponens (inferring B from A → B and A). In Studies 2–4, the information did not license accurate inferences, but some participants made inferential errors: affirming the consequent (inferring A from A → B and B; Study 2) or denying the antecedent (inferring ¬B from A → B and ¬A; Studies 3A, 3B, and 4). Bayesian modeling using ordinal constraints on condition means yielded consistent evidence for the sensitivity of both explicit (self-report) and implicit (IAT and AMP) evaluations to the (correctly or erroneously) inferred truth value of propositions. Together, these data suggest that implicit evaluations, similar to their explicit counterparts, can reflect logical structure.

A Primer on p-Value Thresholds and α-Levels – Two Different Kettles of Fish

It has often been noted that the “null-hypothesis-significance-testing” (NHST) framework is an inconsistent hybrid of Neyman-Pearson’s “hypothesis testing” and Fisher’s “significance testing” that almost inevitably causes misinterpretations. To facilitate a realistic assessment of the potential and the limits of statistical inference, we briefly recall widespread inferential errors and outline the two original approaches of these famous statisticians. Based on the understanding of their irreconcilable perspectives, we propose “going back to the roots” and using the initial evidence in the data in terms of the size and the uncertainty of the estimate for the purpose of statistical inference. Finally, we make six propositions that hopefully contribute to improving the quality of inferences in future research.

Testing for differences between two distributions in the presence of serial correlation using theKolmogorov–Smirnovand Kuiper's tests

AbstractTesting for differences between two states is a staple of climate research, for example, applying a Student'sttest to test for the differences in means. A more general approach is to test for differences in the entire distributions. Increasingly, this latter approach is being used in the context of climate change research where some societal impacts may be more sensitive to changes further from the centre of the distribution. The Kolmogorov–Smirnov (KS) test, probably the most widely‐used method in distributional testing, along with the closely related, but lesser known Kuiper's (KU) test are examined here. These, like most common statistical tests, assume that the data to which they are applied consist of independent observations. Unfortunately, commonly used data such as daily time series of temperature typically violate this assumption due to day‐to‐day autocorrelation. This work explores the consequences of this. Three variants of the KS and KU tests are explored: the traditional approach ignoring autocorrelation, use of an ‘effective sample size’ based on the lag‐1 autocorrelation, and Monte Carlo simulations employing a first order autoregressive model appropriate to a variety of data commonly used in climate science. Results indicate that large errors in inferences are possible when the temporal coherence is ignored. The guidance and materials provided here can be used to anticipate the magnitude of the errors. Bias caused by the errors can be mitigated via easy to use ‘look‐up’ tables or more broadly through application of polynomial coefficients fit to the simulation results.

A Systematic Evaluation of Object Detection Networks for Scientific Plots

Are existing object detection methods adequate for detecting text and visual elements in scientific plots which are arguably different than the objects found in natural images? To answer this question, we train and compare the accuracy of Fast/Faster R-CNN, SSD, YOLO and RetinaNet on the PlotQA dataset with over 220,000 scientific plots. At the standard IOU setting of 0.5, most networks perform well with mAP scores greater than 80% in detecting the relatively simple objects in plots. However, the performance drops drastically when evaluated at a stricter IOU of 0.9 with the best model giving a mAP of 35.70%. Note that such a stricter evaluation is essential when dealing with scientific plots where even minor localisation errors can lead to large errors in downstream numerical inferences. Given this poor performance, we propose minor modifications to existing models by combining ideas from different object detection networks. While this significantly improves the performance, there are still two main issues: (i) performance on text objects which are essential for reasoning is very poor, and (ii) inference time is unacceptably large considering the simplicity of plots. To solve this open problem, we make a series of contributions: (a) an efficient region proposal method based on Laplacian edge detectors, (b) a feature representation of region proposals that includes neighbouring information, (c) a linking component to join multiple region proposals for detecting longer textual objects, and (d) a custom loss function that combines a smooth L1-loss with an IOU-based loss. Combining these ideas, our final model is very accurate at extreme IOU values achieving a mAP of 93.44%@0.9 IOU. Simultaneously, our model is very efficient with an inference time 16x lesser than the current models, including one-stage detectors. Our model also achieves a high accuracy on an extrinsic plot-to-table conversion task with an F1 score of 0.77. With these contributions, we make a definitive progress in object detection for plots and enable further exploration on automated reasoning of plots.

Analysis of Paralogs in Target Enrichment Data Pinpoints Multiple Ancient Polyploidy Events in Alchemilla s.l. (Rosaceae).

Target enrichment is becoming increasingly popular for phylogenomic studies. Although baits for enrichment are typically designed to target single-copy genes, paralogs are often recovered with increased sequencing depth, sometimes from a significant proportion of loci, especially in groups experiencing whole-genome duplication (WGD) events. Common approaches for processing paralogs in target enrichment data sets include random selection, manual pruning, and mainly, the removal of entire genes that show any evidence of paralogy. These approaches are prone to errors in orthology inference or removing large numbers of genes. By removing entire genes, valuable information that could be used to detect and place WGD events is discarded. Here, we used an automated approach for orthology inference in a target enrichment data set of 68 species of Alchemilla s.l. (Rosaceae), a widely distributed clade of plants primarily from temperate climate regions. Previous molecular phylogenetic studies and chromosome numbers both suggested ancient WGDs in the group. However, both the phylogenetic location and putative parental lineages of these WGD events remain unknown. By taking paralogs into consideration and inferring orthologs from target enrichment data, we identified four nodes in the backbone of Alchemilla s.l. with an elevated proportion of gene duplication. Furthermore, using a gene-tree reconciliation approach, we established the autopolyploid origin of the entire Alchemilla s.l. and the nested allopolyploid origin of four major clades within the group. Here, we showed the utility of automated tree-based orthology inference methods, previously designed for genomic or transcriptomic data sets, to study complex scenarios of polyploidy and reticulate evolution from target enrichment data sets.[Alchemilla; allopolyploidy; autopolyploidy; gene tree discordance; orthology inference; paralogs; Rosaceae; target enrichment; whole genome duplication.]

The Wisconsin Card Sorting Test: Split-Half Reliability Estimates for a Self-Administered Computerized Variant

Self-administered computerized assessment has the potential to increase the reach of neuropsychological assessment. The present study reports the first split-half reliability estimates for a self-administered computerized variant of the Wisconsin Card Sorting Test (WCST), which is considered as a gold standard for the neuropsychological assessment of executive functions. We analyzed data from a large sample of young volunteers (N = 375). Split-half reliability estimates for perseveration errors, set-loss errors, and inference errors were all above 0.90. Split-half reliability estimates for response time measures on switch and repeat trials exceeded 0.95. Our results indicated sufficient split-half reliability for a self-administered computerized WCST, paving the way for an advanced digital assessment of executive functions. We discuss potential effects of test formats, administration variants, and sample characteristics on split-half reliability.

Surface of rapidly-rotating neutron stars: Implications to neutron star parameter estimation

The Neutron star Interior Composition Explorer (NICER) is currently observing the x-ray pulse profiles emitted by hot spots on the surface of rotating neutron stars allowing for an inference of their radii with unprecedented precision. A critical ingredient in the pulse profile model is an analytical formula for the oblate shape of the star. These formulas require a fitting over a large ensemble of neutron star solutions, which cover a wide set of equations of state, stellar compactnesses and rotational frequencies. However, this procedure introduces a source of systematic error, as (i) the fits do not describe perfectly the surface of all stars in the ensemble and (ii) neutron stars are described by a single equation of state, whose influence on the surface shape is averaged out during the fitting procedure. Here we perform a first study of this systematic error, finding evidence that it is subdominant relative to the statistical error in the radius inference by NICER. We also find evidence that the formula currently used by NICER can be used in the inference of the radii of rapidly rotating stars, outside of the formula's domain of validity. Moreover, we employ an accurate enthalpy-based method to locate the surface of numerical solutions of rapidly rotating neutron stars and a new highly accurate formula to describe their surfaces. These results can be used in applications that require an accurate description of oblate surfaces of rapidly rotating neutron stars.