Search

• An integrated sampling scheme with adjustable mechanisms is developed. • The OC and ASN functions were derived based on the exact sampling distribution. • Establishment of a nonlinear optimization model to determine optimal plan design. • Several analyses on performance metrics are investigated, compared and discussed. • Development of a web-based application to assist practitioners to obtain optimal plan design. Acceptance sampling plans are statistical quality control methods commonly used to efficiently verify product quality under controlled risks. Recent research has developed the multiple dependent-state sampling plan (MDSP), which incorporates historical lot quality information, and the repetitive group sampling plan (RGSP), which allows for repeat sampling, to enhance the cost-effectiveness of sampling inspections. The modified RGSP (MRGSP) integrates the sampling mechanisms of both MDSP and RGSP. However, investigative analyses have uncovered significant deficiencies in the sampling mechanisms of MDSP and RGSP, with potential problems in MRGSP being even more severe. Therefore, this paper proposes an adjustable MRGSP (AMRGSP) based on unilateral process capability indices to establish a more adaptive and flexible sampling mechanism, reducing the limitations of MRGSP. We derive the operational characteristic function and average sample number function of AMRGSP, and establish a nonlinear optimization model considering Type I and II errors to determine the optimal plan design. Performance comparisons of the proposed AMRGSP with recent sampling plans revealed that the proposed plan offers reliable lot discriminative power and significantly reduces the sample size required for inspection, providing excellent cost-effectiveness. Finally, we evaluate the proposed plan using a practical case study to demonstrate its applicability in practice. [ABSTRACT FROM AUTHOR]

Conversational search systems, such as Google assistant and Microsoft Cortana, enable users to interact with search systems in multiple rounds through natural language dialogues. Evaluating such systems is very challenging, given that any natural language responses could be generated, and users commonly interact for multiple semantically coherent rounds to accomplish a search task. Although prior studies proposed many evaluation metrics, the extent of how those measures effectively capture user preference remain to be investigated. In this article, we systematically meta-evaluate a variety of conversational search metrics. We specifically study three perspectives on thosemetrics: (1) reliability: the ability to detect "actual" performance differences as opposed to those observed by chance; (2) fidelity: the ability to agree with ultimate user preference; and (3) intuitiveness: the ability to capture any property deemed important: adequacy, informativeness, and fluency in the context of conversational search. By conducting experiments on two test collections, we find that the performance of different metrics vary significantly across different scenarios, whereas consistent with prior studies, existing metrics only achieve weak correlation with ultimate user preference and satisfaction. METEOR is, comparatively speaking, the best existing single-turn metric considering all three perspectives. We also demonstrate that adapted session-based evaluation metrics can be used to measure multi-turn conversational search, achieving moderate concordance with user satisfaction. To our knowledge, our work establishes the most comprehensive meta-evaluation for conversational search to date. [ABSTRACT FROM AUTHOR]

Objectives: In China, multiple approaches to calculating EQ-5D utilities are available, including the two EQ-5D-3L (3L2014 and 3L2018) scoring functions, the EQ-5D-5L (5L) scoring function, and the crosswalk function linking the 3L utilities and 5L health states. The study compared utilities derived from them in terms of agreement and discriminative power; and assessed whether the use of different approaches may affect QALY estimation in Chinese type 2 diabetes (T2D) patients. Methods: Cross-sectional data of 289 T2D patients who self-completed both the 5L and 3L questions were used. Agreement were examined using intraclass correlation coefficient (ICC) and Bland–Altman plots. The ability of the EQ-5D utilities in differentiating the patients with and without clinical conditions was evaluated using F-statistics. Their influence on QALY estimation was assessed adopting mean absolute difference (MAD) in utility values between the patients. Results: The ICC values were 0.881 (3L2014-3L2018), 0.958 (5L-c5L2014), and 0.806 (5L-c5L2018). The two 3L utilities and the three 5L utilities had poor agreement at the lower end of utility scale according to Bland–Altman plots. The 3L2018 utilities had lower F-statistics compared to the 3L2014 utilities; the two c5L utilities had larger or similar F-statistics compared to the 5L utilities. The mean MADs were 0.138 (5L), 0.116 (3L2014), 0.115 (c5L2014), 0.055 (c5L2018), and 0.055 (3L2018). Conclusion: The 3L2014 utilities is more discriminative than the 3L2018 utilities; and the two c5L utilities have no worse discriminative power compared with the 5L utilities. The choice of the approach to calculating the EQ-5D utilities is likely to affect QALY estimates. [ABSTRACT FROM AUTHOR]

The evaluation of recommender systems is an area with unsolved questions at several levels. Choosing the appropriate evaluation metric is one of such important issues. Ranking accuracy is generally identified as a prerequisite for recommendation to be useful. Ranking metrics have been adapted for this purpose from the Information Retrieval field into the recommendation task. In this article, we undertake a principled analysis of the robustness and the discriminative power of different ranking metrics for the offline evaluation of recommender systems, drawing from previous studies in the information retrieval field. We measure the robustness to different sources of incompleteness that arise from the sparsity and popularity biases in recommendation. Among other results, we find that precision provides high robustness while normalized discounted cumulative gain offers the best discriminative power. In dealing with cold users, we also find that the geometric mean is more robust than the arithmetic mean as aggregation function over users. [ABSTRACT FROM AUTHOR]

Estimation of the level of durum wheat germplasm genetic diversity is important for its classification and efficient use in breeding programmes. The aim of this study was to assess genetic diversity of durum wheat genotypes developed at the Institute of Field and Vegetable Crops in Novi Sad, Serbia, with 26 morphological characteristics based on the International Union for Protection of New Varieties of Plants guidelines. The Shannon diversity index was used as an indicator of morphological diversity and it ranged from 0.283 for glaucosity of lower side of the flag leaf blade and density of hairiness of uppermost node of the culm, to 0.950 for the ear colour, with the mean value of 0.616, indicating a medium to high level of morphological diversity. On average, the diversity was higher for traits relating to generative organs than for those associated with vegetative plant organs. The 21 morphological characteristics were sufficient to distinguish unique profiles of all durum wheat genotypes. The estimation of varietal diversity and identification of morphological characteristics with the highest discriminative power were done by multiple correspondence analysis. The traits that contributed the most to the distinction of genotypes were the ear colouration, length of beak of the lower glume, lower glume shape, ear length of awns at tip relative to ear length and colour of awns. Morphological characterization using the traits with the highest discriminative power could be a useful complementary method for durum wheat germplasm classification and diversity analysis. [ABSTRACT FROM AUTHOR]

Objective: The main objective of this study is to evaluate the performance of the predictive model (EuroSCORE II) on a Tunisian population to validate its use in our country.Methods: This is a retrospective study of data from 418 adult patients undergoing cardiac surgery with cardiopulmonary bypass between 1 January 2015 and 31 December 2016 in the department of cardiovascular and thoracic surgery of the Sahloul University Hospital of Sousse. The EuroSCORE ΙΙ is calculated using the application validated on the site www.euroscore.org. The performance of the score is evaluated by analyzing its discriminative power by constructing the receiver operating characteristic (ROC) curve and analyzing its calibration using the Hosmer-Lemeshow statistics.Results: The EuroSCORE II shows good discriminative power in our population with an area under the ROC curve more than 0.7 in all study groups (0.864 ± 0.032 for general cardiac surgery, 0.822 ± 0.061 for coronary surgery, 0.864 ± 0.052 for valvular surgery, and 0.900 ± 0.041 for urgent cardiac surgery). The model appears to be calibrated as well by obtaining P values above the statistical significance level of 0.05 (0.638 for general cardiac surgery, 0.543 for coronary surgery, 0.179 for valvular surgery, and 0.082 for urgent cardiac surgery).Conclusion: The EuroSCORE II presents acceptable performance in our population, attested by a good discriminative power and an adequate calibration. [ABSTRACT FROM AUTHOR]

Features act as a key factor in pedestrian detection task. Most widely-used ones like HOG are manually designed and hard to be adaptive, thus now more attention has been paid to the features automatically learned on data. In this paper, a novel approach of learning discriminative features is proposed, addressing two main limitations of the methods in the literature. On one hand, unlike those methods of learning features on low-level pixels, we propose to learn features via a particular sparse coding algorithm enhanced on mid-level image representation, in order to obtain higher-level latent semantics and robustness; On the other hand, those methods usually utilize label information in model training such as deformable part model (DPM) with high computation cost. Instead, we propose to extend the learning process via a maximum margin criterion, in order to better encode discriminative information directly in features by optimizing them to be close to each other if from the same class and far from each other if from different classes. Furthermore, a boosted detection framework rather than the complex DPM is adopted to achieve both high accuracy and efficiency. The proposed approach achieves promising results on several standard pedestrian detection benchmarks. [ABSTRACT FROM AUTHOR]

Soft set theory was originally proposed by Molodtsov in 1999 as a general mathematical tool for dealing with uncertainty. However, it has been pointed out that classical soft set model is not appropriate to deal with imprecise and fuzzy problems. In order to handle these types of problems, some fuzzy extensions of soft set theory are presented, yielding fuzzy soft set theory. As a further research, in this work, we first propose concepts of interval fuzzy sets and interval fuzzy soft sets, define some operations on them and study some of their relevant properties, especially, the dual laws are discussed with respect to difference operation in interval fuzzy soft set theory. We then introduce a revised Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) method and choice value method for interval fuzzy soft set which the weight information is completely unknown. Meanwhile, an analysis of computation complexity is employed, also the discriminative power of two methods are shown. Finally, two illustrative examples are employed to show that they can be successfully applied to problems that contain uncertainties. [ABSTRACT FROM AUTHOR]

Background: In order to evaluate the discrimination performance of an ordinal model for improved disease screening, a new test was proposed where information was obtained across all samples simultaneously. Methods: The ordinal c-index builds upon the volume under the surface methodology without focusing on the accompanying receiver operating characteristic surfaces. However, it can be simplified to an average of pairwise c-indexes. In this paper, a set-based estimate (information was obtained across all samples simultaneously) was proposed by summing all correctly ordered groups. The asymptotic distribution of this proposed estimate was derived using U-statistics. Results: A predictive model was applied using the blood urea nitrogen/creatinine ratio to discriminate stroke in evolution in acute ischemic stroke patients, which could potentially be life-saving in emergency departments. Conclusions: By conducting Monte Carlo simulations, it was concluded that the measure proposed herein is a better choice for clinical use because of the asymmetry of the predicted probabilities of groups. [ABSTRACT FROM AUTHOR]

In this paper we will investigate the gradient utilization in building SIFT (Scale Invariant Feature Transform)-like descriptors for image registration. There are generally two types of gradient information, i.e. gradient magnitude and gradient occurrence, which can be used for building SIFT-like descriptors. We will provide a theoretical analysis on the effectiveness of each of the two types of gradient information when used individually. Based on our analysis, we will propose a novel technique which systematically uses both types of gradient information together for image registration. Moreover, we will propose a strategy to select keypoint matches with a higher discrimination. The proposed technique can be used for both mono-modal and multi-modal image registration. Our experimental results show that the proposed technique improves registration accuracy over existing SIFT-like descriptors. [ABSTRACT FROM AUTHOR]

Objective:The aim of this study was to determine the relationship between hearing loss and speech reception threshold (SRT) in a fixed noise condition using the German Oldenburg sentence test (OLSA).Design:After training with two easily-audible lists of the OLSA, SRTs were determined monaurally with headphones at a fixed noise level of 65 dB SPL using a standard adaptive procedure, converging to 50% speech intelligibility.Study sample:Data was obtained from 315 ears of 177 subjects with hearing losses ranging from − 5 to 90 dB HL pure-tone average (PTA, 0.5, 1, 2, 3 kHz).Results:Two domains were identified with a linear dependence of SRT on PTA. The SRT increased with a slope of 0.094 ± 0.006 dB SNR/dB HL (standard deviation (SD) of residuals = 1.17 dB) for PTAs < 47 dB HL and with a slope of 0.811 ± 0.049 dB SNR/dB HL (SD of residuals = 5.54 dB) for higher PTAs.Conclusion:The OLSA can be applied to subjects with a wide range of hearing losses. With 65 dB SPL fixed noise presentation level the SRT is determined by listening in noise for PTAs < ∼47 dB HL, and above it is determined by listening in quiet. [ABSTRACT FROM PUBLISHER]

Purpose: To compare the discriminative power of the index scores of EQ-5D-5L (5L) and EQ-5D-3L (3L) in diabetes patients in China. Methods: A consecutive sample of type 2 diabetes mellitus (T2DM) patients in the clinics self-completed the two versions of EQ-5D. The 3L index score was obtained from the Chinese 3L value set; the 5L index score was predicted from the 3L index score using an interim scoring. Relative efficiency (RE) of the 5L and 3L index scores was calculated to compare their ability in differentiating between T2DM patients with and without one of ten clinical conditions. The efficiency of the 5L and 3L health state classification systems was assessed using the Shannon index ( H′) and in terms of ceiling effects. Results: A total of 289 T2DM patients participated in this study. The 5L score was systematically lower than the 3L score for T2DM patients with and without a condition (range −0.36 to −0.06). The 5L score exhibited higher discriminative power in nine of ten conditions, with the mean RE value being 1.92. 5L had higher H′ values than 3L in all the five EQ-5D dimensions: mobility (1.14 vs. 0.70), self-care (0.44 vs. 0.33), usual activities (0.72 vs. 0.47), pain/discomfort (1.58 vs. 1.10), and anxiety/depression (1.03 vs. 0.67). The overall ceiling effects decreased from 56.7 % (3L) to 36.7 % (5L). Conclusion: The 5L index score is more discriminative than the 3L index score in T2DM patients and therefore is preferable for use in this population. [ABSTRACT FROM AUTHOR]

In this study, several feature combinations are studied to analyse their relevance for online signature verification. Different time functions associated with the signing process are analysed in order to provide some insight on their actual discriminative power. This analysis could also help forensic handwriting experts (FHEs) to further understand the signatures and the writer's behaviour. Among the different feature combinations analysed, a set of features which seems to be relevant for signature analysis by FHEs is particularly considered. The feasibility of developing a system which could complement the FHEs work is evaluated. Two different approximations of the analysed time functions are proposed, one based on the Legendre polynomials and another based on the wavelet decomposition. The coefficients in these orthogonal series expansions of the time functions are used as features to model them. Two different signature styles are considered, namely, Western and Chinese, of one of the most recent publicly available signature databases. The experimental results are promising, in particular for the features that seem to be relevant for the FHEs, since the obtained verification error rates are comparable with the ones reported in the state‐of‐the‐art over the same datasets. [ABSTRACT FROM AUTHOR]

Purpose: The short form six dimensions (SF-6D) are derived from the SF-36 Health Survey with the intention that item data of the latter are often converted to a preference value, which was subsequently used in economic evaluations of interventions. The aim was to compare the equivalence and sensitivity of health preference values derived from the SF-36/SF-12 Health Surveys to that measured directly by the SF-6D for chronic hepatitis B (CHB) patients. Methods: This was a secondary analysis of the SF-6D and SF-36 data from a baseline sample of 589 patients with CHB infection with different stages of liver diseases. Degree of agreement (equivalence) between direct-measured and derived SF-6D values was determined using spearman correlation and intra-class correlation. Sensitivity and discriminative power of different SF-6D values were compared by standardized effect size and relative efficiency (RE) statistics. Results: Significant differences in the direct-measured or derived SF-6D preference values were found between CHB groups. Degree of agreement between SF-6D values was satisfactory. Direct-measured SF-6D was the most efficient, followed by SF-12-derived and the SF-36-derived was the least, based on the standardized effect size and the RE statistics. Sensitivity and discriminative power of direct-measured SF-6D were superior to derived SF-6D among people with different CHB health states. Conclusions: Although direct-measured and derived SF-6D preference values had satisfactory sensitivity in discriminating between CHB groups, direct-measured SF-6D is the most sensitive and preferable method of obtaining health preference. [ABSTRACT FROM AUTHOR]

Background/purpose Visual characteristics such as color and shape of skin lesions play an important role in the diagnostic process. In this contribution, we quantify the discriminative power of such attributes using an information theoretical approach. Methods We estimate the probability of occurrence of each attribute as a function of the skin diseases. We use the distribution of this probability across the studied diseases and its entropy to define the discriminative power of the attribute. The discriminative power has a maximum value for attributes that occur (or do not occur) for only one disease and a minimum value for those which are equally likely to be observed among all diseases. Results Verrucous surface, red and brown colors, and the presence of more than 10 lesions are among the most informative attributes. A ranking of attributes is also carried out and used together with a naive Bayesian classifier, yielding results that confirm the soundness of the proposed method. Conclusion proposed measure is proven to be a reliable way of assessing the discriminative power of dermatological attributes, and it also helps generate a condensed dermatological lexicon. Therefore, it can be of added value to the manual or computer-aided diagnostic process. [ABSTRACT FROM AUTHOR]

The Hough forest method is an effective method for object detection in ground-shot images that has received increasing research attention. However, this method lacks the ability to detect objects with arbitrary orientations. This largely constrains the method from being used in detecting geospatial objects from remotely sensed (RS) images since geospatial objects can have many different orientations. In order to achieve rotation invariance and compensate the associated loss of discriminative power, this paper presents a novel color-enhanced rotation-invariant Hough forest (CRIHF) method for detecting geospatial objects in RS images. In our method, we propose to train a Pose-Estimation-based Rotation-invariant Texton Forest (PE-RTF) which first uses dominant gradient orientations to align local image patches. The orientations are then jointly used with coordinates in Hough voting to detect object position. In order to increase discriminative power, Texton Forest is used in codebook generation. Moreover, theoretically sound color-invariant gradients are employed. By rotating split functions rather than image patches in the RTF and sparsely accumulating Hough votes on grid points, computational times can be reduced by two orders of magnitude. The evaluation of the CRIHF method on a data set containing 525 airplanes and a second data set containing 68 residential buildings shows that our method is rotation invariant and robust. The detector achieves around 90% recall rate on both data sets. Experiments also show that our method is noise resistant and can achieve a decent detection performance at a high level (30%) of “salt and pepper” impulsive noise. [ABSTRACT FROM AUTHOR]

Objective: To compare results on the everyday sentence test 'FIST', the new closed-set sentence test 'FrMatrix', and the digit triplet screening test 'FrDigit3'. Design: First, the FrMatrix was developed and normative values were obtained. Subsequently, speech reception thresholds (SRTs) for the three types of tests were gathered at four study centers representing different geographic regions in Belgium and France. Study sample: Fifty-seven normal-hearing listeners took part in the normative study of the FrMatrix, and 118 subjects, with a wide range of hearing thresholds, participated in the comparative study. Results: Homogenizing the individual words of the FrMatrix with regard to their intelligibility resulted in a reference SRT of −6.0 (±0.6) dB SNR and slope at the SRT of 14.0 %/dB. The within-subject variability was only 0.4 dB. Comparison of the three tests showed high correlations between the SRTs mutually (>0.81). The FrMatrix had the highest discriminative power, both in stationary and in fluctuating noise. For all three tests, differences across the participating study centers were small and not significant. Conclusions: The FIST, the FrMatrix, and the FrDigit3 provide similar results and reliably evaluate speech recognition performance in noise both in normal-hearing and hearing-impaired listeners. [ABSTRACT FROM AUTHOR]

Growing numbers of candidate varieties, decrease of their variability for morphological traits, and internationalization of the national list all contribute to excessive increase of the trial costs, thus creating the need for the improvement of current variety evaluation procedures, especially regards their distinctness, Uniformity, and Stability (DUS) component. Due to rapid advancement in molecular techniques, the use of molecular markers in DUS testing as a complement to, or replacement of, morphological observations became the subject of great interest in scientific studies, and consequently topic for discussion within International Union for the Protection of New Varieties of Plants (UPOV). In order to explore the potential of simple sequence repeat (SSR) markers for distinctness tests, present study involved set of 41 maize inbred lines that were scored for 32 DUS characters prescribed by UPOV and genotyped at 28 SSR loci. Results were largely in favor of the use of molecular markers, revealing or confirming their already known advantages over morphological markers like better consistency with the pedigree, and relatively higher discriminative power. However, their integration into DUS testing protocols still depends upon resolving of several important issues. [ABSTRACT FROM AUTHOR]

Feature selection is important and necessary for disease classification and prediction using high‐dimensional gene expression data. A hybrid method integrating sparse representation with a two‐sample statistical t‐test to construct features from high‐throughput microarray data is presented. The approach takes account of gene interaction and reduces the variable dimension by sparse linear combination, as well as considers the discriminative power of genes using component regression. Under the recurrent independence rule for classification, the experiment results on real data demonstrate the improvements of this hybrid technique over conventional methods. [ABSTRACT FROM AUTHOR]

Probing food experience or liking through verbal ratings has its shortcomings. We compare explicit ratings to a range of (neuro)physiological and behavioral measures with respect to their performance in distinguishing drinks associated with different emotional experience. Seventy participants tasted and rated the valence and arousal of eight regular drinks and a "ground truth" high-arousal, low-valence vinegar solution. The discriminative power for distinguishing between the vinegar solution and the regular drinks was highest for sip size, followed by valence ratings, arousal ratings, heart rate, skin conductance level, facial expression of "disgust," pupil diameter, and Electroencephalogram (EEG) frontal alpha asymmetry. Within the regular drinks, a positive correlation was found between rated arousal and heart rate, and a negative correlation between rated arousal and Heart Rate Variability (HRV). Most physiological measures showed consistent temporal patterns over time following the announcement of the drink and taking a sip. This was consistent over all nine drinks, but the peaks were substantially higher for the vinegar solution than for the regular drinks, likely caused by emotion. Our results indicate that implicit variables have the potential to differentiate between drinks associated with different emotional experiences. In addition, this study gives us insight into the physiological temporal response patterns associated with taking a sip. [ABSTRACT FROM AUTHOR]

Background: Neuroinflammation has gained increasing attention as a potential contributing factor in Alzheimer’s disease (AD) pathology. A clinical cerebrospinal fluid biomarker capable of monitoring this process during the course of the disease has yet to emerge, chiefly owing to contradictory research findings. In this study, we sought to clarify the utility of inflammatory biomarkers in diagnostic procedures of AD in three steps: (1) to screen for proteins that are robustly detectable in cerebrospinal fluid; (2) based on this analysis, to explore any associations between the analytically robust markers and salient pathological features of AD; and (3) to determine the discriminative power of these markers in the clinical diagnosis of AD. Methods: From a total of 46 proteins, 15 that were robustly detectable in cerebrospinal fluid were identified. A subsequent analysis of these markers in a cohort of 399 patients (nondemented subjects, patients with mild cognitive impairment [MCI], and patients with AD, supplemented by smaller cohorts of other diseases) was conducted. Fluid biomarker data were related to AD pathology and neuropsychological markers and adjusted for confounders such as age, sex, apolipoprotein E genotype, and biobank storage time. Results: Cerebrospinal fluid levels of C-reactive protein and soluble TREM2 differed between nondemented subjects, patients with MCI, or patients with AD and were associated with amyloid and tau pathology. Several markers were associated with tau pathology only or with other neurodegenerative diseases. Correlations between neuropsychological performance and inflammatory markers were weak, but they were most prominent in AD and for the most challenging cognitive tests. All investigated covariates had significant influence, with varying effects across the markers. Still, none of the markers achieved discriminative power of more than 70% to distinguish between patient groups defined by clinical or neuropathological categories. Conclusions: Basic analytical considerations proved indispensable for this type of study because only one-third of the tested markers were robustly detectable in cerebrospinal fluid. Detectable inflammatory protein markers were associated in multiple ways with AD pathology. Yet, even significantly associated markers were not powerful enough in terms of effect strength, sensitivity, and specificity, and hence they were not suited for direct use in clinical diagnostic practice. Targets other than those most commonly considered in this field of research might provide results with better clinical applicability. [ABSTRACT FROM AUTHOR]

A vein recognition method using the local block pattern is presented. After the phase information of a vein image is extracted by a group of Gabor filters and local XOR pattern operators, the phase coding image is divided into some even non‐overlap blocks. These local block patterns can alleviate the effect of noises and enhance the discriminative power. Each block is divided further into some sub‐blocks, and Fisher's linear discriminant is applied here to reduce the dimensionality of the proposed descriptor and improve the performance at the same time. Finally, these histograms of blocks are concatenated to form the descriptor of the whole vein image. The proposed method is evaluated on the hand vein image database and the HK PolyU's database. The results show that the local block pattern strategy plays a crucial role in the vein recognition method. When the sizes of the block and the sub‐block are right, the method can achieve an error equation rate of 0.88 and 0.13% on the two databases, respectively. [ABSTRACT FROM AUTHOR]

Several object-oriented cohesion metrics have been proposed in the literature. These metrics aim to measure the relationship between class members, namely, methods and attributes. Different metrics use different models to represent the connectivity pattern of cohesive interactions (CPCI) between class members. Most of these metrics are normalized to allow for easy comparison of the cohesion of different classes. However, in some cases, these metrics obtain the same cohesion values for different classes that have the same number of methods and attributes but different CPCIs. This leads to incorrectly considering the classes to be the same in terms of cohesion, even though their CPCIs clearly indicate that the degrees of cohesion are different. We refer to this as a lack of discrimination anomaly (LDA) problem. In this paper, we list and discuss cases in which the LDA problem exists, as expressed through the use of 16 cohesion metrics. In addition, we empirically study the frequent occurrence of the LDA problem when the considered metrics are applied to classes in five open source Java systems. Finally, we propose a metric and a simulation-based methodology to measure the discriminative power of cohesion metrics. The discrimination metric measures the probability that a cohesion metric will produce distinct cohesion values for classes with the same number of attributes and methods but different CPCIs. A highly discriminating cohesion metric is more desirable because it exhibits a lower chance of incorrectly considering classes to be cohesively equal when they have different CPCIs. [ABSTRACT FROM AUTHOR]