Likelihood Ratio Approach Research Articles

Evidential value of polymeric materials-chemometric tactics for spectral data compression combined with likelihood ratio approach.

Polymers have become a ubiquitous element of our culture. Therefore, these materials may play an important role in forensic investigations, serving as mute witnesses of occurrences such as car accidents. In this study, the possibilities provided by the likelihood ratio (LR) approach to estimate the evidential value of observed similarities and differences, and to discriminate among NIR spectral data originating from polypropylene automotive parts and household items, were investigated. Since the construction of LR models requires the introduction of only a few variables, the main objective was to reduce the dimensionality of registered spectra, which are characterised by over a thousand variables. The applied strategy was based on compression of NIR signals using discrete wavelet transform (DWT) followed by use of the SELECT algorithm for the selection and decorrelation of the most informative DWT coefficients. Selected features eventually served as an input for LR models. The performance of the developed models was assessed by measuring the rates of false positive and false negative answers as well as by applying an empirical cross entropy approach. Despite relatively small databases of polymeric objects, both univariate and multivariate LR models showed acceptable performances. The latter, however, gave the most satisfactory results, as it enabled successful discrimination of compared samples and delivered the lowest error rates. In addition, in order to verify the potential of NIR spectroscopy, the obtained results were compared with those obtained after application of the proposed tactics to the FTIR data, which is a well-established method in the forensic sphere.

Maximized likelihood ratio tests for functional localization in fMRI

fMRI localizer tasks are often used to define subject specific functional regions of interest (fROIs) that contain the relevant features for subsequent analyses. fROIs are typically small and show large interindividual differences in extent and effect size. As statistical testing procedures focus on con- trolling false positives, this may lead to an ad-hoc adjustment of thresholding in some individuals. The promising likelihood ratio (LR) testing approach for fMRI (Kang et al. 2015) provides simultaneous control of both false positives and negatives by contrasting evidence in favor of true activation against evidence in favor of the null hypothesis. The authors propose to estimate the expected alternative by a percentile (e.g. 95th) across the voxels of an effect size map. However, in the context of fROIs, pre-defined observed percentiles may induce inconsistent activation across subjects. In this study we show the potential of a maximized LR approach (Bickel, 2012) for this particular application. The maximum LR is calculated over the same interval of functionally relevant alternatives for all subjects, enabling consistent localization of the fROIs in both subjects with low levels and high levels of general activity.

Evaluation of direct and indirect ethanol biomarkers using a likelihood ratio approach to identify chronic alcohol abusers for forensic purposes

The detection of direct ethanol metabolites, such as ethyl glucuronide (EtG) and fatty acid ethyl esters (FAEEs), in scalp hair is considered the optimal strategy to effectively recognize chronic alcohol misuses by means of specific cut-offs suggested by the Society of Hair Testing. However, several factors (e.g. hair treatments) may alter the correlation between alcohol intake and biomarkers concentrations, possibly introducing bias in the interpretative process and conclusions.125 subjects with various drinking habits were subjected to blood and hair sampling to determine indirect (e.g. CDT) and direct alcohol biomarkers. The overall data were investigated using several multivariate statistical methods. A likelihood ratio (LR) approach was used for the first time to provide predictive models for the diagnosis of alcohol abuse, based on different combinations of direct and indirect alcohol biomarkers. LR strategies provide a more robust outcome than the plain comparison with cut-off values, where tiny changes in the analytical results can lead to dramatic divergence in the way they are interpreted. An LR model combining EtG and FAEEs hair concentrations proved to discriminate non-chronic from chronic consumers with ideal correct classification rates, whereas the contribution of indirect biomarkers proved to be negligible. Optimal results were observed using a novel approach that associates LR methods with multivariate statistics. In particular, the combination of LR approach with either Principal Component Analysis (PCA) or Linear Discriminant Analysis (LDA) proved successful in discriminating chronic from non-chronic alcohol drinkers. These LR models were subsequently tested on an independent dataset of 43 individuals, which confirmed their high efficiency. These models proved to be less prone to bias than EtG and FAEEs independently considered. In conclusion, LR models may represent an efficient strategy to sustain the diagnosis of chronic alcohol consumption and provide a suitable gradation to support the judgment.

Local Ion Signatures (LIS) for the examination of comprehensive two-dimensional gas chromatography applied to fire debris analysis

Forensic examination of fire debris evidence is a notoriously difficult analytical task due to the complexity and variability of sample composition. The use of comprehensive two-dimensional gas chromatography with mass spectrometry detection (GC×GC–MS) allows the coupling of orthogonal retention mechanisms and therefore a high peak capacity.We demonstrate recent innovations in combining chemometric techniques for data reduction and feature selection, with evaluation of the evidence for forensic questions pertaining to the detection and subsequent classification of ignitable liquid residue (ILR) in fire debris samples. Chromatograms are divided into non-overlapping spatially delimited regions; for each of these regions a Local Ion Signature (LIS) is computed by summing the intensities, per nominal mass/charge over all points contained within each region. This yields a reduced feature space representing the original data as a set of consolidated ion traces. Subsequent feature selection is performed by evaluating the individual efficacy of each feature using a univariate score-based likelihood ratio (LR) approach for discriminating between pairs of same or different type samples. The retained features are used to model each ILR class using linear discriminant analysis (LDA).Results are demonstrated for 155 arson samples containing a diversity of substrate compounds and several known ignitable liquids. ILR detection is performed at 84% accuracy with fewer than 1% false positives followed by subsequent classification. Likelihood ratio distributions are presented referring to both detection and classification tasks.

Performance of confidence intervals for the population size in capture–recapture experiment under inverse sampling with replacement

ABSTRACTLui [Five confidence intervals of the closed population size in the capture-recapture problem under inverse sampling with replacement. Biom J. 2004;46:474–480] considered five confidence intervals for the closed population size in a two-sample capture–recapture experiment under inverse sampling with replacement in the recapture phase. The results of his Monte Carlo study indicated that the exact confidence intervals and those based on -approximation perform very well. In this paper, we consider three other methods of interval estimation including the bootstrap, likelihood ratio and Jeffreys prior approaches. A Monte Carlo simulation is carried out to evaluate the performance of these intervals together with those based on existing methods in terms of the coverage probability, error rates and standardized average length. Our results show that confidence intervals based on Wald statistics, logarithmic transformation, and bootstrap methods are inappropriate, having coverage probabilities less than the desired nominal level. Also, the exact confidence intervals and those based on -approximation are not invariant with respect to the proportion of marked individuals in the capture phase, say p. When p is chosen to be small-to-moderate, the likelihood ratio method is preferred, since it gives confidence intervals with shorter length from among all methods that provide the coverage probability close to the desired nominal level. Overall, the Jeffreys method appears to be more robust than other competitors, providing intervals with nearest coverage probabilities to the desired nominal level and with balanced non-coverage rates.

Detection of spatial change points in the mean and covariances of multivariate simultaneous autoregressive models.

In this paper, we propose a test procedure to detect change points of multidimensional autoregressive processes. The considered process differs from typical applied spatial autoregressive processes in that it is assumed to evolve from a predefined center into every dimension. Additionally, structural breaks in the process can occur at a certain distance from the predefined center. The main aim of this paper is to detect such spatial changes. In particular, we focus on shifts in the mean and the autoregressive parameter. The proposed test procedure is based on the likelihood-ratio approach. Eventually, the goodness-of-fit values of the estimators are compared for different shifts. Moreover, the empirical distribution of the test statistic of the likelihood-ratio test is obtained via Monte Carlo simulations. We show that the generalized Gumbel distribution seems to be a suitable limiting distribution of the proposed test statistic. Finally, we discuss the detection of lung cancer in computed tomography scans and illustrate the proposed test procedure.

On the enhanced detectability of GPS anomalous behavior with relative entropy

A standard receiver autonomous integrity monitoring (RAIM) technique for the global positioning system (GPS) has been dedicated to provide an integrity monitoring capability for safety-critical GPS applications, such as in civil aviation for the en-route (ER) through non-precision approach (NPA) or lateral navigation (LNAV). The performance of the existing RAIM method, however, may not meet more stringent aviation requirements for availability and integrity during the precision approach and landing phases of flight due to insufficient observables and/or untimely warning to the user beyond a specified time-to-alert in the event of a significant GPS failure. This has led to an enhanced RAIM architecture ensuring stricter integrity requirement by greatly decreasing the detection time when a satellite failure or a measurement error has occurred. We thus attempted to devise a user integrity monitor which is capable of identifying the GPS failure more rapidly than a standard RAIM scheme by incorporating the RAIM with the relative entropy, which is a likelihood ratio approach to assess the inconsistence between two data streams, quite different from a Euclidean distance. In addition, the delay-coordinate embedding technique needs to be considered and preprocessed to associate the discriminant measure obtained from the RAIM with the relative entropy in the new RAIM design. In simulation results, we demonstrate that the proposed user integrity monitor outperforms the standard RAIM with a higher level of detection rate of anomalies which could be hazardous to the users in the approach or landing phase and is a very promising alternative for the detection of deviations in GPS signal. The comparison also shows that it enables to catch even small anomalous gradients more rapidly than a typical user integrity monitor.

Hybrid approach combining chemometrics and likelihood ratio framework for reporting the evidential value of spectra

Many chemometric tools are invaluable and have proven effective in data mining and substantial dimensionality reduction of highly multivariate data. This becomes vital for interpreting various physicochemical data due to rapid development of advanced analytical techniques, delivering much information in a single measurement run. This concerns especially spectra, which are frequently used as the subject of comparative analysis in e.g. forensic sciences. In the presented study the microtraces collected from the scenarios of hit-and-run accidents were analysed. Plastic containers and automotive plastics (e.g. bumpers, headlamp lenses) were subjected to Fourier transform infrared spectrometry and car paints were analysed using Raman spectroscopy. In the forensic context analytical results must be interpreted and reported according to the standards of the interpretation schemes acknowledged in forensic sciences using the likelihood ratio approach. However, for proper construction of LR models for highly multivariate data, such as spectra, chemometric tools must be employed for substantial data compression. Conversion from classical feature representation to distance representation was proposed for revealing hidden data peculiarities and linear discriminant analysis was further applied for minimising the within-sample variability while maximising the between-sample variability. Both techniques enabled substantial reduction of data dimensionality. Univariate and multivariate likelihood ratio models were proposed for such data. It was shown that the combination of chemometric tools and the likelihood ratio approach is capable of solving the comparison problem of highly multivariate and correlated data after proper extraction of the most relevant features and variance information hidden in the data structure.

Evaluating a subset of ancestry informative SNPs for discriminating among Southwest Asian and circum-Mediterranean populations

Many different published sets of single nucleotide polymorphisms (SNPs) and/or insertion-deletion polymorphisms (InDels) can serve as ancestry informative markers (AIMs) to distinguish among continental regions of the world. For a focus on Southwest Asian ancestry we chose to start with the Kidd Lab panel of 55 ancestry-informative SNPs (AISNPs) because it already provided good global reference data (FROG-kb: frog.med.yale.edu) in a set of 73 population samples distinguishing at least 8 biogeographic clusters of populations. This panel serves as a good first tier ancestry panel. We are now interested in identifying region-specific second tier panels for more refined distinction among populations within each of the global regions. We have begun studying the global region centered on Southwest Asia and the region encompassing the Mediterranean Sea. We have incorporated 10 populations from North Africa, Turkey and Iran and included 31 of the original 73 populations and eleven 1000 Genomes Phase3 populations for a total of 3129 individuals from 52 populations, all typed for the 55 AISNPs. We have then identified the subset of the 55 AISNPs that are most informative for this region of the world using Heatmap, Fst, and Informativeness analyses to eliminate those SNPs essentially redundant or providing no information among populations in this region, reducing the number of SNPs to 32. STRUCTURE and PCA analyses show the remaining 32 SNPs identify the North African cluster and appropriately include the Turkish and Iranian samples with the Southwest Asian cluster. These markers provide the basis for building an improved, optimized panel of AISNPs that provides additional information on differences among populations in this part of the world. The data have also allowed an examination of the accuracy of the ancestry inference based on 32 SNPs for the newly studied populations from this region. The likelihood ratio approach to ancestry inference embodied in FROG-kb provides highly significant population assignments within one order of magnitude for each individual in the Turkish, Iranian, and Tunisian populations.

Assessing evidentiary value in fire debris analysis by chemometric and likelihood ratio approaches

Results are presented from support vector machine (SVM), linear and quadratic discriminant analysis (LDA and QDA) and k-nearest neighbors (kNN) methods of binary classification of fire debris samples as positive or negative for ignitable liquid residue. Training samples were prepared by computationally mixing data from ignitable liquid and substrate pyrolysis databases. Validation was performed on an unseen set of computationally mixed (in silico) data and on fire debris from large-scale research burns. The probabilities of class membership were calculated using an uninformative (equal) prior and a likelihood ratio was calculated from the resulting class membership probabilities. The SVM method demonstrated a high discrimination, low error rate and good calibration for the in silico validation data; however, the performance decreased significantly for the fire debris validation data, as indicated by a significant increase in the error rate and decrease in the calibration. The QDA and kNN methods showed similar performance trends. The LDA method gave poorer discrimination, higher error rates and slightly poorer calibration for the in silico validation data; however the performance did not deteriorate for the fire debris validation data.

Automated Surface Defect Detection Using High-Density Data

State-of-the-art measurement technologies, such as 3D laser scanners, provide new opportunities for knowledge discovery and development of quality control (QC) strategies for complex manufacturing systems. These technologies can rapidly provide millions of data points to represent a manufactured part's surface. The resulting high-density (HD) datasets have a great potential to be used for inspecting parts for surface and feature abnormalities. The current use of these datasets for part inspection can be divided into two main categories: (1) extracting feature parameters, which does not complement the nature of these datasets as it wastes valuable data and (2) an ad hoc inspection process, where a visual representation of the data is manually analyzed, which tends to suffer from slow, inefficient, and inconsistent inspection results. To overcome these deficiencies, this paper proposes an adaptive generalized likelihood ratio (AGLR) technique to automate the surface defect inspection process using HD data. This paper presents the performance results of the proposed AGLR approach with respect to the probability of detecting varying size and magnitude defects in addition to the probability of false alarms. In addition, a formal approach for designing an optimal AGLR inspection system is proposed. Finally, simulation results are presented and analyzed to showcase the performance gains of the AGLR approach versus a more traditional generalized likelihood ratio (GLR) approach.

Improvement in detection of minor alleles in next generation sequencing by base quality recalibration.

BackgroundMinor allele detection in very high coverage sequence data (>1000X) has many applications such as detecting mtDNA heteroplasmy, somatic mutations in cancer or tumors, SNP calling in pool sequencing, etc., where reads with low frequency are not necessarily sequence error but may instead convey biological information. However, the suitability of common base quality recalibration tools for such applications has not been investigated in detail.ResultsWe show that the widely used tool GATK BaseRecalibration has several limitations in minor allele detection. First, GATK IndelRealignment fails to work if the sequence coverage is above a certain level since it then becomes computationally infeasible. Second, the accuracy of the base quality largely depends on the database of known SNPs as the control, which limits the ability of de novo minor allele detection. Third, GATK reduces the base quality of sequence errors at the cost of reducing scores for true minor alleles. To overcome these limitations, we present a novel approach called SEGREG, which applies segmented regression to control sequences (e.g. phiX174 DNA) spiked into a sequencing run. Based on simulations SEGREG improves both the accuracy of base quality scores and the detection of minor alleles. We further investigate sequence error and recalibration parameters by applying a Logarithm Likelihood Ratio (LLR) approach to SEGREG recalibrated base quality scores for phiX174 DNA sequenced to very high coverage, and for mtDNA genome sequences previously analyzed for heteroplasmic variants.ConclusionsOur results suggest that SEGREG improves base recalibration without suffering the limitations discussed above, and the LLR approach benefits from SEGREG in identifying more true minor alleles, while avoiding false positives from sequencing error.Electronic supplementary materialThe online version of this article (doi:10.1186/s12864-016-2463-2) contains supplementary material, which is available to authorized users.

Robust multipixel matched subspace detection with signal-dependent background power

A modified matched subspace detector (MSD) has been recently proposed for detecting a barely discernible object in an additive Gaussian background clutter using a single pixel in a sequence of digital images. In contrast to this detector designed for the subpixel object, we developed a generalized likelihood ratio approach to the detection of a multipixel object of unknown shape, size, and position in an additive signal-dependent Gaussian background and noise. The proposed detector modifies the MSD by adding the additional term proportional to the square of the difference between the background variances under two statistical hypotheses. The performances of these detectors are evaluated for the example scenario of two multipixel floating objects on the agitated sea surface. The crucial characteristic of the proposed detector is that prior knowledge of the target size, shape, and position is not required. Computer simulation and experimental results have shown that the proposed detector outperforms the MSD, especially in the case of weak and poorly contrasted objects of unknown shape, size, and position.

Evaluation of infrared spectra analyses using a likelihood ratio approach: A practical example of spray paint examination

Depending on the forensic disciplines and on the analytical techniques used, Bayesian methods of evaluation have been applied both as a two-step approach (first comparison, then evaluation) and as a continuous approach (comparison and evaluation in one step). However in order to use the continuous approach, the measurements have to be reliably summarized as a numerical value linked to the property of interest, which occurrence can be determined (e.g., refractive index measurement of glass samples).For paint traces analyzed by Fourier transform infrared spectroscopy (FTIR) however, the statistical comparison of the spectra is generally done by a similarity measure (e.g., Pearson correlation, Euclidean distance). Although useful, these measures cannot be directly associated to frequencies of occurrence of the chemical composition (binders, extenders, pigments). The continuous approach as described above is not possible, and a two-step evaluation, 1) comparison of the spectra and 2) evaluation of the results, is therefore the common practice reported in most of the laboratories. Derived from a practical question that arose during casework, a way of integrating the similarity measure between spectra into a continuous likelihood ratio formula was explored. This article proposes the use of a likelihood ratio approach with the similarity measure of infrared spectra of spray paints based on distributions of sub-populations given by the color and composition of spray paint cans. Taking into account not only the rarity of paint composition, but also the “quality” of the analytical match provides a more balanced evaluation given source or activity level propositions. We will demonstrate also that a joint statistical–expertal methodology allows for a more transparent evaluation of the results and makes a better use of current knowledge.

On the Univariate Representation of BEKK Models with Common Factors

AbstractSimple low order multivariate GARCH models imply marginal processes with a lot of persistence in the form of high order lags. This is not what we find in many situations however, where parsimonious univariate GARCH(1,1) models for instance describe quite well the conditional volatility of some asset returns. In order to explain this paradox, we show that in the presence of common GARCH factors, parsimonious univariate representations can result from large multivariate models generating the conditional variances and conditional covariances/correlations. The diagonal model without any contagion effects in conditional volatilities gives rise to similar conclusions though. Consequently, after having extracted a block of assets representing some form of parsimony, remains the task of determining if we have a set of independent assets or instead a highly dependent system generated with a few factors. To investigate this issue, we first evaluate a reduced rank regressions approach for squared returns that we extend to cross-returns. Second we investigate a likelihood ratio approach, where under the null the matrix parameters have a reduced rank structure. It emerged that the latter approach has quite good properties enabling us to discriminate between a system with seemingly unrelated assets (e.g. a diagonal model) and a model with few common sources of volatility.

An Adjusted Likelihood Ratio Approach Analysing Distribution of Food Products to Assist the Investigation of Foodborne Outbreaks.

In order to facilitate foodborne outbreak investigations there is a need to improve the methods for identifying the food products that should be sampled for laboratory analysis. The aim of this study was to examine the applicability of a likelihood ratio approach previously developed on simulated data, to real outbreak data. We used human case and food product distribution data from the Norwegian enterohaemorrhagic Escherichia coli outbreak in 2006. The approach was adjusted to include time, space smoothing and to handle missing or misclassified information. The performance of the adjusted likelihood ratio approach on the data originating from the HUS outbreak and control data indicates that the adjusted approach is promising and indicates that the adjusted approach could be a useful tool to assist and facilitate the investigation of food borne outbreaks in the future if good traceability are available and implemented in the distribution chain. However, the approach needs to be further validated on other outbreak data and also including other food products than meat products in order to make a more general conclusion of the applicability of the developed approach.

Detection of seizures in intracranial EEG: UPenn and Mayo Clinic's Seizure Detection Challenge.

A system for detection of seizures in intracranial EEG is presented that is based on a combination of generative, discriminative and hybrid approaches. We present a methodology to effectively benefit from the advantages each classifier offers. In particular, Gaussian mixture models, Support Vector Machines, hybrid likelihood ratio and Gaussian supervector approaches are developed and combined for the task. This system participated in the UPenn and Mayo Clinic's Seizure Detection Challenge, ranking in the top 5 of over 200 participants. The drawbacks of the proposed method with respect to the winning solutions are critically assessed.

Application of a likelihood ratio approach in solving a comparison problem of Raman spectra recorded for blue automotive paints

Evidential values of Raman spectra recorded for solid and metallic blue car paints were evaluated using visual comparison as well as a statistical approach, i.e. likelihood ratio (LR) test supported by Empirical Cross Entropy (ECE) results. Raman spectra were obtained using a Renishaw inVia spectrometer equipped with a confocal Leica microscope and a near infrared semiconductor laser (785 nm) as an excitation source.When a visual comparison was performed for solid paints, in 17.7% of comparisons (53 pairs out of 300) the samples were indistinguishable, whereas for metallic paints this value was 8.5% of such comparisons (37 pairs out of 435). Comparing the spectra using an LR approach was based on variables being the areas under the most significant Raman spectra bands of pigments present in the analysed samples. Proposed LR models delivered low false positive and false negative rates (for many models lower than 10%), and the ECE plots confirmed that their performance was much better than visual comparison. Copyright © 2015 John Wiley & Sons, Ltd.

Development of non-alcoholic fatty liver disease scoring system among adult medical check-up patients: a large cross-sectional and prospective validation study

BackgroundNon-alcoholic fatty liver disease (NAFLD) is the most common liver disease in the community. However, NAFLD remains undiagnosed in most people with limited access to imaging facilities in most developing countries.ObjectiveTo examine the prevalence of NAFLD and to develop the risk scoring model for predicting the presence of NAFLD among adult medical check-up patients.MethodA large prospective cross-sectional study was conducted among medical check-up patients who underwent transabdominal ultrasound examination between January and December 2013 in Medistra Hospital, Jakarta. Data were obtained from the patients’ medical records. Logistic regression analyses were undertaken to identify the best combination of risk factors for predicting fatty liver using the backward (likelihood ratio) approach. The adjusted odds ratio and 95% confidence interval were estimated using the logistic regression coefficient. The prediction model was assessed using the receiver operating characteristic curve and the Hosmer–Lemeshow goodness-of-fit test and was validated on a new, prospective cohort. Statistical analysis was done using SPSS version 17.ResultsA total of 1,054 cases was included in this study. Fatty liver was present in 538 (51.0%) patients. Bivariate analyses found associations among fatty liver and several risk factors. Six risk factors were incorporated to build the final prediction model. All scores were summed up to obtain the total score. A probability equation was developed by applying linear regression analysis on the total score. The prediction model had good diagnostic performance with an area under the receiver operating characteristic curve =0.833 (95% confidence interval =0.809–0.857). The Hosmer–Lemeshow goodness-of-fit P-value was 0.232, which indicated the appropriateness of the logistic regression model to predict fatty liver. On the validation set, the scoring system proved to be moderately accurate and can potentially be applied to larger population setting.ConclusionThe presence of fatty liver in NAFLD patients can be predicted using our proposed fatty liver scoring system.

Interpretation of FTIR spectra of polymers and Raman spectra of car paints by means of likelihood ratio approach supported by wavelet transform for reducing data dimensionality.

The problem of interpretation of common provenance of the samples within the infrared spectra database of polypropylene samples from car body parts and plastic containers as well as Raman spectra databases of blue solid and metallic automotive paints was under investigation. The research involved statistical tools such as likelihood ratio (LR) approach for expressing the evidential value of observed similarities and differences in the recorded spectra. Since the LR models can be easily proposed for databases described by a few variables, research focused on the problem of spectra dimensionality reduction characterised by more than a thousand variables. The objective of the studies was to combine the chemometric tools easily dealing with multidimensionality with an LR approach. The final variables used for LR models' construction were derived from the discrete wavelet transform (DWT) as a data dimensionality reduction technique supported by methods for variance analysis and corresponded with chemical information, i.e. typical absorption bands for polypropylene and peaks associated with pigments present in the car paints. Univariate and multivariate LR models were proposed, aiming at obtaining more information about the chemical structure of the samples. Their performance was controlled by estimating the levels of false positive and false negative answers and using the empirical cross entropy approach. The results for most of the LR models were satisfactory and enabled solving the stated comparison problems. The results prove that the variables generated from DWT preserve signal characteristic, being a sparse representation of the original signal by keeping its shape and relevant chemical information.