Discrimination Statistics Research Articles

Determining the Success of Economic Sanctions

This article applies the social-network conception of national power, derived from the social-network conception of the international system structure, to economic sanctions, one of the most widely studied empirical phenomena in international relations. The empirical analyses of economic sanctions presented here find that sanctions cases with disproportional structural-network power between sender and target were far less likely to be successful and those with the target state possessing high structural-network power were far more likely to be successful. The evidence from nonparametric model discrimination statistics and information criteria measures shows that the sanctions models with new social-network power measures have greater explanatory power than or statistically outperform those with old attributional power measures, such as the Correlates of War index and GNP. 一些政治学者认为强调农业和农民的“乡村情结”越来越失去其政治和社会意义，但并没有多少实证研究支持这一说法。为此，作者就人们对农业和农民的态度做了探讨性调查。调查结果显示，人们的态度符合“乡村情结式微”论，但会因年龄、位置、出生国、投票意向而有差异。不过，仍存在一种将乡村情结归诸农民的倾向，有对农民生产方式的强烈支持、以及对农业于国家未来重要性的坚信。令人惊奇的是，与过去三十年的市场自由主义相左，对政府之扶助农业的支持可谓强烈。

Validation of use of the International Consultation on Incontinence Questionnaire‐Urinary Incontinence‐Short Form (ICIQ‐UI‐SF) for impairment rating: A transversal retrospective study of 120 patients

A urinary incontinence impairment rating must be a highly accurate, non-invasive exploration of the condition using International Classification of Functioning (ICF)-based assessment tools. The objective of this study was to identify the best evaluation test and to determine an impairment rating model of urinary incontinence. In performing a cross-sectional study comparing successive urodynamic tests using both the International Consultation on Incontinence Questionnaire-Urinary Incontinence-Short Form (ICIQ-UI-SF) and the 1-hr pad-weighing test in 120 patients, we performed statistical likelihood ratio analysis and used logistic regression to calculate the probability of urodynamic incontinence using the most significant independent predictors. Subsequently, we created a template that was based on the significant predictors and the probability of urodynamic incontinence. The mean ICIQ-UI-SF score was 13.5 ± 4.6, and the median pad test value was 8 g. The discrimination statistic (receiver operating characteristic) described how well the urodynamic observations matched the ICIQ-UI-SF scores (under curve area (UDA):0.689) and the pad test data (UDA: 0.693). Using logistic regression analysis, we demonstrated that the best independent predictors of urodynamic incontinence were the patient's age and the ICIQ-UI-SF score. The logistic regression model permitted us to construct an equation to determine the probability of urodynamic incontinence. Using these tools, we created a template to generate a probability index of urodynamic urinary incontinence. Using this probability index, relative to the patient and to the maximum impairment of the whole person (MIWP) relative to urinary incontinence, we were able to calculate a patient's permanent impairment.

Evidence of community structure in Biomedical Research Grant Collaborations

Recent studies have clearly demonstrated a shift towards collaborative research and team science approaches across a spectrum of disciplines. Such collaborative efforts have also been acknowledged and nurtured by popular extramurally funded programs including the Clinical Translational Science Award (CTSA) conferred by the National Institutes of Health. Since its inception, the number of CTSA awardees has steadily increased to 60 institutes across 30 states. One of the objectives of CTSA is to accelerate translation of research from bench to bedside to community and train a new genre of researchers under the translational research umbrella. Feasibility of such a translation implicitly demands multi-disciplinary collaboration and mentoring. Networks have proven to be convenient abstractions for studying research collaborations. The present study is a part of the CTSA baseline study and investigates existence of possible community-structure in Biomedical Research Grant Collaboration (BRGC) networks across data sets retrieved from the internally developed grants management system, the Automated Research Information Administrator (ARIA) at the University of Arkansas for Medical Sciences (UAMS).Fastgreedy and link-community community-structure detection algorithms were used to investigate the presence of non-overlapping and overlapping community-structure and their variation across years 2006 and 2009. A surrogate testing approach in conjunction with appropriate discriminant statistics, namely: the modularity index and the maximum partition density is proposed to investigate whether the community-structure of the BRGC networks were different from those generated by certain types of random graphs.Non-overlapping as well as overlapping community-structure detection algorithms indicated the presence of community-structure in the BRGC network. Subsequent, surrogate testing revealed that random graph models considered in the present study may not necessarily be appropriate generative mechanisms of the community-structure in the BRGC networks. The discrepancy in the community-structure between the BRGC networks and the random graph surrogates was especially pronounced at 2009 as opposed to 2006 indicating a possible shift towards team-science and formation of non-trivial modular patterns with time. The results also clearly demonstrate presence of inter-departmental and multi-disciplinary collaborations in BRGC networks. While the results are presented on BRGC networks as a part of the CTSA baseline study at UAMS, the proposed methodologies are as such generic with potential to be extended across other CTSA organizations. Understanding the presence of community-structure can supplement more traditional network analysis as they’re useful in identifying research teams and their inter-connections as opposed to the role of individual nodes in the network. Such an understanding can be a critical step prior to devising meaningful interventions for promoting team-science, multi-disciplinary collaborations, cross-fertilization of ideas across research teams and identifying suitable mentors. Understanding the temporal evolution of these communities may also be useful in CTSA evaluation.

Application of a model-based depth discriminator to data from the REP11 experiment

We address application of a passive, model-based depth discriminator to data from the REP11 experiment. The method is based on a mode subspace approach (Premus, 2007) which uses environmental information along with a normal mode based acoustic simulation to predict the propagating mode structure. This mode space can be divided into subspaces representing the lower and higher order modes. Sufficient aperture yields orthogonal and linearly independent subspaces and a linear algebraic process yields orthogonalized subspaces with reduced aperture. Received data is then projected onto these subspaces and a discrimination statistic is formed. This work examines the application of this process to data from the REP11 experiment in terms of performance of the discriminator over different sets of data and levels of environmental knowledge. Work sponsored by ONR Undersea Signal Processing.

Nowcasting of kinetic energy of hail precipitation using radar

The detection of hail precipitation generated by a storm is a complicated task due to the limited spatial extension and the space-time irregularity of impacts generated on the ground. Some of the most extensive methods to create climatology of these impacts are observer networks or hailpad networks. Both methods are affected by numerous inconveniences, overall when it is necessary to work with an extensive area, in which it is necessary to maintain an operating network that has numerous maintenance costs.In this sense, there are numerous works done that have developed different models with the objective of detecting hail precipitation using meteorological radar. Some of these methods use discriminant statistic techniques that, through the combination of different radar parameters, can achieve very satisfactory results. On the other hand, it would be very interesting to know not only the probability of hail, but also some of the characteristics of the hailstones precipitated, such as the number or their kinetic energy, since these parameters are directly related to the damage generated in infrastructures and/or crops.The estimation of kinetic energy of hail precipitation using meteorological radar has caught the interest of some authors. In our case, we used the databases obtained by hailpad networks and the databases of C-band and S-Band radar to build an algorithm to estimate the vertical component of kinetic energy produced by a hail precipitation. In order to carry out this study, data on hail was gathered and analyzed from the hailpad networks in the province of Zaragoza (in the north-east of Spain) and the province of Mendoza (in Argentina, close to the Andes range on the border with Chile). These are two geographically distant regions, but which share a common characteristic: a high frequency of storms with hail precipitation, mainly during the summer months (Sánchez et al., 2009a).In order to compile the database, we have established two categories of kinetic energy: hail with low energy (<20Jm−2) and hail with high energy (>20Jm−2). With this information, we have looked to establish a differentiation between hail precipitation that hardly produces damage and, on the contrary, that which does (Dessens et al., 2007). Once these two categories were established, we constructed a logistic function that establishes a correspondence between radar variables and the two categories of kinetic energy.The results show great uncertainty in determining kinetic energy using C-band radar. However, for the S-band radar the results have shown that the probability of detection is 85.7% with a FAR of 14.3% and an explained variance of 61.2%. This result allows us, with corresponding caution, to make a first estimation of the areas in which a hailstorm could produce damage.With the objective of making a meteorological interpretation of the results obtained (achieved only using statistic criteria), we have previously done an Principal Components Analysis (PCA).

Large-Deviation Results for Discriminant Statistics of Gaussian Locally Stationary Processes

This paper discusses the large-deviation principle of discriminant statistics for Gaussian locally stationary processes. First, large-deviation theorems for quadratic forms and the log-likelihood ratio for a Gaussian locally stationary process with a mean function are proved. Their asymptotics are described by the large deviation rate functions. Second, we consider the situations where processes are misspecified to be stationary. In these misspecified cases, we formally make the log-likelihood ratio discriminant statistics and derive the large deviation theorems of them. Since they are complicated, they are evaluated and illustrated by numerical examples. We realize the misspecification of the process to be stationary seriously affecting our discrimination.

Validation of Default Probabilities

AbstractWell-performing default predictions show good discrimination and calibration. Discrimination is the ability to separate defaulters from nondefaulters. Calibration is the ability to make unbiased forecasts. I derive novel discrimination and calibration statistics to verify forecasts expressed in terms of probability under dependent observations. The test statistics’ asymptotic distributions can be derived in analytic form. Not accounting for cross correlation can result in the rejection of actually well-performing predictions, as shown in an empirical application. I demonstrate that forecasting errors must be serially uncorrelated. As a consequence, my multiperiod tests are statistically consistent.

Comparisons of established risk prediction models for cardiovascular disease: systematic review

Objective To evaluate the evidence on comparisons of established cardiovascular risk prediction models and to collect comparative information on their relative prognostic performance.Design Systematic review of comparative predictive model studies.Data...

Predicting the risk of chronic kidney disease in the UK: an evaluation of QKidney® scores using a primary care database

Chronic kidney disease is a major health concern that, if left untreated, may progress to end-stage kidney failure (ESKF). Identifying individuals at an increased risk of kidney disease and who might benefit from a therapeutic or preventive intervention is an important challenge. To evaluate the performance of the QKidney® scores for predicting 5-year risk of developing moderate-severe kidney disease and ESKF in an independent UK cohort of patients from general practice records. Prospective cohort study to evaluate the performance of two risk scores for kidney disease in 364 practices from the UK, contributing to The Health Improvement Network (THIN) database. Data were obtained from 1.6 million patients registered with a general practice surgery between 1 January 2002 and 1 July 2008, aged 35-74 years, with 43,186 incident cases of moderate-severe kidney disease and 2663 incident cases of ESKF. This is the first recorded evidence of moderate-severe chronic kidney and ESKF as recorded in general practice records. The results from this independent and external validation of QKidney scores indicate that both scores showed good performance data for both moderate-severe kidney disease and ESKF, on a large cohort of general practice patients. Discrimination and calibration statistics were better for models including serum creatinine; however, there were considerable amounts of missing data for serum creatinine. QKidney scores both with and without serum creatinine were well calibrated. QKidney scores have been shown to be useful tools to predict the 5-year risk of moderate-severe kidney disease and ESKF in the UK.

Discriminating between the Weibull and log-normal distributions for Type-II censored data

Log-normal and Weibull distributions are the two most popular distributions for analysing lifetime data. In this paper, we consider the problem of discriminating between the two distribution functions. It is assumed that the data are coming either from log-normal or Weibull distributions and that they are Type-II censored. We use the difference of the maximized log-likelihood functions, in discriminating between the two distribution functions. We obtain the asymptotic distribution of the discrimination statistic. It is used to determine the probability of correct selection in this discrimination process. We perform some simulation studies to observe how the asymptotic results work for different sample sizes and for different censoring proportions. It is observed that the asymptotic results work quite well even for small sizes if the censoring proportions are not very low. We further suggest a modified discrimination procedure. Two real data sets are analysed for illustrative purposes.

Comparison of clinical and angiographic prognostic risk scores in patients with acute coronary syndromes: Analysis from the Acute Catheterization and Urgent Intervention Triage StrategY (ACUITY) trial

Several prognostic risk scores have been developed for patients with coronary artery disease, but their comparative use in patients with non-ST-segment elevation acute coronary syndromes (NSTEACS) undergoing percutaneous coronary intervention (PCI) has not been examined. We therefore investigated the accuracy of the Synergy Between PCI With Taxus and Cardiac Surgery (SYNTAX) score, Clinical Synergy Between PCI With Taxus and Cardiac Surgery score (CSS), New Risk Stratification (NERS) score (NERS), Age, Creatinine, Ejection Fraction (ACEF) score, Global Registry for Acute Coronary Events (GRACE) score, and Thrombolysis in Myocardial Infarction (TIMI) score for risk assessment of 1-year mortality, cardiac mortality, myocardial infarction, target vessel revascularization, and stent thrombosis in patients with NSTEACS undergoing PCI. The 6 scores were determined in 2,094 patients with NSTEACS treated with PCI enrolled in the angiographic substudy of the ACUITY trial. The prognostic accuracy of the 6 scores was assessed using the c statistic for discrimination and the Hosmer-Lemeshow test for calibration. The index of separation and net reclassification improvement (NRI) were also determined. Scores incorporating clinical and angiographic variables (CSS and NERS) showed the best tradeoff between discrimination and calibration for most end points, with the best discrimination for all end points and good calibration for most of them. The CSS had the best index of separation for most ischemic endpoints and displayed an NRI for cardiac death and myocardial infarction (MI) compared to the other scores, whereas NERS displayed an NRI for all-cause death and target vessel revascularization. The 3 scores-CSS, NERS, and SYNTAX-were the only scores to have both good discrimination and calibration for cardiac mortality. In patients with NSTEACS undergoing PCI, risk scores incorporating clinical and angiographic variables had the highest predictive accuracy for a broad spectrum of ischemic end points.

Abstract 147: The Association Between Hematocrit and Mortality among Ischemic Stroke Patients

Introduction: Anemia is associated with higher mortality among patients with such non-stroke vascular conditions as heart failure and myocardial infarction. Less is known regarding the relationship between anemia and mortality among patients with acute ischemic stroke. Methods: Medical records were abstracted for a sample of 3965 veterans from 131 Veterans Health Administration (VHA) facilities who were admitted for a confirmed diagnosis of ischemic stroke (fiscal year 2007). Hematocrit (Hct) values from 24-hours of admission were categorized into 6-tiers (≤27%, 28-32%, 33-37%, 38-42%, 43-47%, ≥48%). We excluded patients with: female gender (n=95), incomplete Hct data (n=94), thrombolysis (n=32), and inconsistent death dates (n=6). We used multivariate logistic regression to examine the relationship between anemia and in-hospital, 30-day, 60-day and one-year mortality using multivariate logistic regression models for each time point, adjusting for age, NIHSS, comorbidity (including pneumonia), and Acute Physiology and Chronic Health Evaluation (APACHE)-III scores. The discrimination (c-statistics) and calibration (Hosmer-Lemeshow goodness of fit [HLGOF]) statistics were generated to gauge model performance and fit. Results: Approximately 2.1% of the N=3750 patients presented with Hcts ≤27%, 6.2% were 28-32%, 17.9% were 33-37%, 36.4% were 38-42%, 28.2% were 43-47%, and 9.1% were ≥48%. Adjusted mortality odds at all time points were 2.5 to 3.5 times higher for those with ≤Hct 27% (p values &lt; 0.013 for in-hospital and 30-day mortality; p values at 6 months and one year were 0.002 and 0.001, respectively). Mortality risk at 6 months and 1 year showed a significant and dose-response relationship to Hct for all Hct groups &lt;38%. High Hcts were independently associated only with in-hospital mortality and only in those with Hct ≥48 (OR 2.9, p=0.004). Models performed well across time points (C=0.813, HLGOF=0.9684 [in-hospital]; C=0.832, HLGOF=0.8186 [30-day]; C=0.863, HLGOF=0.7307 [60-day]; C=0.880, HLGOF=0.4313 [one-year]). Conclusions: Even a moderate level of anemia is independently associated with an increased risk of death during the first year following acute ischemic stroke. Very low or very high Hct is associated with early post-stroke mortality. Further work is required to evaluate whether interventions that treat anemia, its complications and underlying etiologies may also reduce post-stroke mortality.

Association of C-Reactive Protein With Cardiovascular Disease Mortality According to Diabetes Status

OBJECTIVEC-reactive protein (CRP) is associated with the risk of cardiovascular disease (CVD); whether the effects are modified by diabetes status still is unclear. This study investigated these issues and assessed the added value of CRP to predictions.RESEARCH DESIGN AND METHODSParticipants were drawn from representative samples of adults living in England and Scotland. Cox proportional hazards regression models were used to relate baseline plasma CRP with all-cause and CVD mortality during follow-up in men and women with and without diabetes. The added value of CRP to the predictions was assessed through c-statistic comparison and relative integrated discrimination improvement.RESULTSA total of 25,979 participants (4.9% with diabetes) were followed for a median of 93 months, during which period there were 2,767 deaths (957 from CVD). CRP (per SD loge) was associated with a 53% (95% CI 43–64) and 43% (38–49) higher risk of cardiovascular and all-cause mortality, respectively. These associations were log linear and did not differ according to diabetes status (both P ≥ 0.08 for interaction), sex, and other risk factors. Adding CRP to conventional risk factors improved predictions overall and separately by diabetes status but not for CVD mortality, although such improvements only were marginal based on several discrimination statistics.CONCLUSIONSThe association between CRP and CVD was similar across diabetes status, and the effects are broadly similar across levels of other conventional risk factors.

Discovery and validation of gene classifiers for endocrine-disrupting chemicals in zebrafish (danio rerio).

BackgroundDevelopment and application of transcriptomics-based gene classifiers for ecotoxicological applications lag far behind those of biomedical sciences. Many such classifiers discovered thus far lack vigorous statistical and experimental validations. A combination of genetic algorithm/support vector machines and genetic algorithm/K nearest neighbors was used in this study to search for classifiers of endocrine-disrupting chemicals (EDCs) in zebrafish. Searches were conducted on both tissue-specific and tissue-combined datasets, either across the entire transcriptome or within individual transcription factor (TF) networks previously linked to EDC effects. Candidate classifiers were evaluated by gene set enrichment analysis (GSEA) on both the original training data and a dedicated validation dataset.ResultsMulti-tissue dataset yielded no classifiers. Among the 19 chemical-tissue conditions evaluated, the transcriptome-wide searches yielded classifiers for six of them, each having approximately 20 to 30 gene features unique to a condition. Searches within individual TF networks produced classifiers for 15 chemical-tissue conditions, each containing 100 or fewer top-ranked gene features pooled from those of multiple TF networks and also unique to each condition. For the training dataset, 10 out of 11 classifiers successfully identified the gene expression profiles (GEPs) of their targeted chemical-tissue conditions by GSEA. For the validation dataset, classifiers for prochloraz-ovary and flutamide-ovary also correctly identified the GEPs of corresponding conditions while no classifier could predict the GEP from prochloraz-brain.ConclusionsThe discrepancies in the performance of these classifiers were attributed in part to varying data complexity among the conditions, as measured to some degree by Fisher’s discriminant ratio statistic. This variation in data complexity could likely be compensated by adjusting sample size for individual chemical-tissue conditions, thus suggesting a need for a preliminary survey of transcriptomic responses before launching a full scale classifier discovery effort. Classifier discovery based on individual TF networks could yield more mechanistically-oriented biomarkers. GSEA proved to be a flexible and effective tool for application of gene classifiers but a similar and more refined algorithm, connectivity mapping, should also be explored. The distribution characteristics of classifiers across tissues, chemicals, and TF networks suggested a differential biological impact among the EDCs on zebrafish transcriptome involving some basic cellular functions.

Discharge destination following lower limb fracture: Development of a prediction model to assist with decision making

BackgroundAccurate prediction of the likelihood of discharge to inpatient rehabilitation following lower limb fracture made on admission to hospital may assist patient discharge planning and decrease the burden on the hospital system caused by delays in decision making. AimsTo develop a prognostic model for discharge to inpatient rehabilitation. MethodIsolated lower extremity fracture cases (excluding fractured neck of femur), captured by the Victorian Orthopaedic Trauma Outcomes Registry (VOTOR), were extracted for analysis. A training data set was created for model development and validation data set for evaluation. A multivariable logistic regression model was developed based on patient and injury characteristics. Models were assessed using measures of discrimination (C-statistic) and calibration (Hosmer–Lemeshow (H–L) statistic). ResultsA total of 1429 patients met the inclusion criteria and were randomly split into training and test data sets. Increasing age, more proximal fracture type, compensation or private fund source for the admission, metropolitan location of residence, not working prior to injury and having a self-reported pre-injury disability were included in the final prediction model. The C-statistic for the model was 0.92 (95% confidence interval (CI) 0.88, 0.95) with an H–L statistic of χ2=11.62, p=0.17. For the test data set, the C-statistic was 0.86 (95% CI 0.83, 0.90) with an H–L statistic of χ2=37.98, p<0.001. ConclusionA model to predict discharge to inpatient rehabilitation following lower limb fracture was developed with excellent discrimination although the calibration was reduced in the test data set. This model requires prospective testing but could form an integral part of decision making in regards to discharge disposition to facilitate timely and accurate referral to rehabilitation and optimise resource allocation.

Efficiencies for the statistics of size discrimination

Different laboratories have achieved a consensus regarding how well human observers can estimate the average orientation in a set of N objects. Such estimates are not only limited by visual noise, which perturbs the visual signal of each object's orientation, they are also inefficient: Observers effectively use only √N objects in their estimates (e.g., S. C. Dakin, 2001; J. A. Solomon, 2010). More controversial is the efficiency with which observers can estimate the average size in an array of circles (e.g., D. Ariely, 2001, 2008; S. C. Chong, S. J. Joo, T.-A. Emmanouil, & A. Treisman, 2008; K. Myczek & D. J. Simons, 2008). Of course, there are some important differences between orientation and size; nonetheless, it seemed sensible to compare the two types of estimate against the same ideal observer. Indeed, quantitative evaluation of statistical efficiency requires this sort of comparison (R. A. Fisher, 1925). Our first step was to measure the noise that limits size estimates when only two circles are compared. Our results (Weber fractions between 0.07 and 0.14 were necessary for 84% correct 2AFC performance) are consistent with the visual system adding the same amount of Gaussian noise to all logarithmically transduced circle diameters. We exaggerated this visual noise by randomly varying the diameters in (uncrowded) arrays of 1, 2, 4, and 8 circles and measured its effect on discrimination between mean sizes. Efficiencies inferred from all four observers significantly exceed 25% and, in two cases, approach 100%. More consistent are our measurements of just-noticeable differences in size variance. These latter results suggest between 62 and 75% efficiency for variance discriminations. Although our observers were no more efficient comparing size variances than they were at comparing mean sizes, they were significantly more precise. In other words, our results contain evidence for a non-negligible source of late noise that limits mean discriminations but not variance discriminations.

Geochemical characterization of ochre from central coastal British Columbia, Canada

The use of ochre and its presence in archaeological contexts is well documented archaeologically, ethnographically and historically. Elemental analysis of ochre is becoming increasingly common as a method for identifying ancient quarrying and mining practices, and for identifying patterns of resource use, trade and exchange. As this type of work has become more common it has become apparent that two critical issues need to be considered: (1) the need for systematic identification of trends and patterns in ochre geochemistry using appropriate statistical analyses, and, (2) consideration of geographic and geologic scale in relation to ochre chemistry. Because ochre is a heterogeneous material relative to other raw material types, it is necessary to apply multivariate and discriminant statistics to differentiate geochemical groups within a sample set. However, caution must be taken when interpreting statistical results at face value. Local geologic and geomorphologic conditions play a significant role in ochre chemistry and need to be taken into account when interpreting analytical results. The purpose of this project was to determine if geochemically distinct groups of ochre could be distinguished through neutron activation analysis (NAA) of archaeological and geological ochres from the central coast of British Columbia, Canada. The results show that it is possible to satisfy the provenance postulate and to differentiate chemical groups through elemental characterization and using multivariate statistical methods.

Predicting risk of osteoporotic and hip fracture in the United Kingdom: prospective independent and external validation of QFractureScores

Objective To evaluate the performance of the QFractureScores for predicting the 10 year risk of osteoporotic and hip fractures in an independent UK cohort of patients from general practice records.Design...

Network Power and Militarized Conflicts

This study argues that a state’s power is best conceptualized by considering how a state interacts with all other states in different networks within the international system. This social network explanation for national power is applied to militarized conflicts, one of the most widely studied empirical phenomena in international relations, focusing on their power explanations. The empirical analysis of militarized conflicts at the dyadic level produces findings that strongly support power preponderance theory over balance of power theory. The evidence from nonparametric model discrimination statistics and information criteria measures also shows that our conflict models with new social network power measures have greater explanatory power than or statistically outperform models relying on attributional power measures, such as Correlates of War index and Gross National Product (GNP).

Fasting glucose cutoff point: where does the risk terminate? Tehran lipid and glucose study

The aim of this study is to determine safe fasting plasma glucose (FPG) levels. We included data on 5,960 individuals aged ≥20 years at baseline with at least one follow-up examination. Diabetes was ascertained in accordance with American Diabetes Association criteria, using standard 2-h post-challenge plasma glucose test. Multivariate restricted cubic splines Weibull regression was implemented for interval-censored survival data on incident diabetes. We used Harrell's C statistic for discrimination, Nam-D'Agostino χ(2) for calibration, and Royston's R(2) for variations in the outcome explained by models. During a 6-year median follow-up, 369 incident cases of diabetes were ascertained. Family history of diabetes, systolic blood pressure, waist-to-height ratio, and triglyceride-to-high-density lipoprotein cholesterol ratio, independent of FPG and each other remained associated with incident diabetes. The cubic splines model achieved good calibration (χ(2) = 12.1) and discrimination (C = 0.828) and explained 75% of variation in the time until incident diabetes. A J-shaped FPG-diabetes relationship was observed. Descending arm of the dose-response relationship curve corresponded to increasing FPG levels up to 4.0 mmol l(-1), where it started increasing. The risk of incident diabetes decreased with decreasing levels of FPG down to 4.0 mmol l(-1), where the risk stopped decreasing. Multivariate-adjusted risk of incident diabetes was zero at FPG = 5.05 mmol l(-1). Although currently there is no firm evidence to show that early intervention among individuals with the elevated FPG levels could prevent or delay onset of diabetes, individuals with FPG levels below 5.05 mmol l(-1) could be safely reassured about their near-term risk of developing incident diabetes and screened on a less frequent basis.