Penalized Regression Models Research Articles

Exact model comparisons in the plausibility framework

Plausibility is a formalization of exact tests for parametric models and generalizes procedures such as Fisher’s exact test. The resulting tests are based on cumulative probabilities of the probability density function and evaluate consistency with a parametric family while providing exact control of the α level for finite sample size. Model comparisons are inefficient in this approach. We generalize plausibility by incorporating weighing which allows to perform model comparisons. We show that one weighing scheme is asymptotically equivalent to the likelihood ratio test (LRT) and has finite sample guarantees for the test size under the null hypothesis unlike the LRT. We confirm theoretical properties in simulations that mimic the data set of our data application. We apply the method to a retinoblastoma data set and demonstrate a parent-of-origin effect.Weighted plausibility also has applications in high-dimensional data analysis and P-values for penalized regression models can be derived. We demonstrate superior performance as compared to a data-splitting procedure in a simulation study. We apply weighted plausibility to a high-dimensional gene expression, case-control prostate cancer data set.We discuss the flexibility of the approach by relating weighted plausibility to targeted learning, the bootstrap, and sparsity selection.

Using DNA to predict behaviour problems from preschool to adulthood.

One goal of the DNA revolution is to predict problems in order to prevent them. We tested here if the prediction of behaviour problems from genome-wide polygenic scores (GPS) can be improved by creating composites across ages and across raters and by using a multi-GPS approach that includes GPS for adult psychiatric disorders as well as for childhood behaviour problems. Our sample included 3,065 genotyped unrelated individuals from the Twins Early Development Study who were assessed longitudinally for hyperactivity, conduct, emotional problems, and peer problems as rated by parents, teachers, and children themselves. GPS created from 15 genome-wide association studies were used separately and jointly to test the prediction of behaviour problems composites (general behaviour problems, externalising, and internalising) across ages (from age 2 to 21) and across raters in penalised regression models. Based on the regression weights, we created multi-trait GPS reflecting the best prediction of behaviour problems. We compared GPS prediction to twin heritability using the same sample and measures. Multi-GPS prediction of behaviour problems increased from <2% of the variance for observed traits to up to 6% for cross-age and cross-rater composites. Twin study estimates of heritability, although to a lesser extent, mirrored patterns of multi-GPS prediction as they increased from <40% to 83%. The ability of GPS to predict behaviour problems can be improved by using multiple GPS, cross-age composites and cross-rater composites, although the effect sizes remain modest, up to 6%. Our approach can be used in any genotyped sample to create multi-trait GPS predictors of behaviour problems that will be more predictive than polygenic scores based on a single age, rater, or GPS.

Bayesian LASSO Regression with Asymmetric Error in High Dimensional

One of the most critical discussions in regression models is the selection of the optimal model, by identifying critical explanatory variables and negligible variables and more easily express the relationship between the response variable and explanatory variables. Given the limitations of selecting variables in classical methods, such as stepwise selection, it is possible to use penalized regression methods. One of the penalized regression models is the Lasso regression model, in which it is assumed that errors follow a normal distribution. In this paper, we introduce the Bayesian Lasso regression model with an asymmetric distribution error and the high dimensional setting. Then, using the simulation studies and real data analysis, the performance of the proposed model's performance is discussed.

A comparison of prediction approaches for identifying prodromal Parkinson disease

Identifying people with Parkinson disease during the prodromal period, including via algorithms in administrative claims data, is an important research and clinical priority. We sought to improve upon an existing penalized logistic regression model, based on diagnosis and procedure codes, by adding prescription medication data or using machine learning. Using Medicare Part D beneficiaries age 66–90 from a population-based case-control study of incident Parkinson disease, we fit a penalized logistic regression both with and without Part D data. We also built a predictive algorithm using a random forest classifier for comparison. In a combined approach, we introduced the probability of Parkinson disease from the random forest, as a predictor in the penalized regression model. We calculated the receiver operator characteristic area under the curve (AUC) for each model. All models performed well, with AUCs ranging from 0.824 (simplest model) to 0.835 (combined approach). We conclude that medication data and random forests improve Parkinson disease prediction, but are not essential.

Maternal Body Mass Index, Early-Pregnancy Metabolite Profile, and Birthweight.

ContextMaternal prepregnancy body mass index (BMI) has a strong influence on gestational metabolism, but detailed metabolic alterations are unknown.ObjectiveFirst, to examine the associations of maternal prepregnancy BMI with maternal early-pregnancy metabolite alterations. Second, to identify an early-pregnancy metabolite profile associated with birthweight in women with a higher prepregnancy BMI that improved prediction of birthweight compared to glucose and lipid concentrations.Design, Setting, and ParticipantsPrepregnancy BMI was obtained in a subgroup of 682 Dutch pregnant women from the Generation R prospective cohort study.Main Outcome MeasuresMaternal nonfasting targeted amino acids, nonesterified fatty acid, phospholipid, and carnitine concentrations measured in blood serum at mean gestational age of 12.8 weeks. Birthweight was obtained from medical records.ResultsA higher prepregnancy BMI was associated with 72 altered amino acids, nonesterified fatty acid, phospholipid and carnitine concentrations, and 6 metabolite ratios reflecting Krebs cycle, inflammatory, oxidative stress, and lipid metabolic processes (P-values < 0.05). Using penalized regression models, a metabolite profile was selected including 15 metabolites and 4 metabolite ratios based on its association with birthweight in addition to prepregnancy BMI. The adjusted R2 of birthweight was 6.1% for prepregnancy BMI alone, 6.2% after addition of glucose and lipid concentrations, and 12.9% after addition of the metabolite profile.ConclusionsA higher maternal prepregnancy BMI was associated with altered maternal early-pregnancy amino acids, nonesterified fatty acids, phospholipids, and carnitines. Using these metabolites, we identified a maternal metabolite profile that improved prediction of birthweight in women with a higher prepregnancy BMI compared to glucose and lipid concentrations.

Predicting life with a permanent end colostomy: A prospective study on function, bother and acceptance

The factors that influence a patient's experience of a colostomy are not known. The aim of this study was to characterise stoma function, stoma-related bother and acceptance among patients operated for rectal cancer and to investigate if there were any preoperative personal factors with predictive impact on long-term stoma-related bother. The QoLiRECT (Quality of Life in RECTal cancer) study is a prospective multicentre study of patients with rectal cancer. This was a subgroup analysis of patients with a permanent colostomy with a 2-year follow-up. Penalised regression models with shrinkage estimation were used to predict the 1-and 2-year bother using baseline data. The predictive value and the importance of the included variables were evaluated using bootstrap resampling techniques. A total of 379 patients were included. Overall stoma acceptance was high and a majority of patients were not bothered by their stoma; 77% and 83% at 1 and 2years, respectively. The subgroup of patients with stoma-related bother had a high prevalence of difficulties, especially fear of leakage, and a low stoma acceptance in daily life. Both clinical and personal factors were associated with stoma-related bother. The most important factors were quality of life and physical health, but the prediction accuracy was low. Stoma-related bother was associated with overall stoma dysfunction. As stoma-related bother is a multifactorial problem, it was not possible to predict which patients will experience stoma-related bother. It is therefore of importance to prevent stoma-related symptoms and optimise stoma function to reduce long-term bother and increase stoma acceptance.

Delta band activity contributes to the identification of command following in disorder of consciousness

The overt or covert ability to follow commands in patients with disorders of consciousness is considered a sign of awareness and has recently been defined as cortically mediated behaviour. Despite its clinical relevance, the brain signatures of the perceptual processing supporting command following have been elusive. This multimodal study investigates the temporal spectral pattern of electrical brain activity to identify features that differentiated healthy controls from patients both able and unable to follow commands. We combined evidence from behavioural assessment, functional neuroimaging during mental imagery and high-density electroencephalography collected during auditory prediction, from 21 patients and 10 controls. We used a penalised regression model to identify command following using features from electroencephalography. We identified seven well-defined spatiotemporal signatures in the delta, theta and alpha bands that together contribute to identify DoC subjects with and without the ability to follow command, and further distinguished these groups of patients from controls. A fine-grained analysis of these seven signatures enabled us to determine that increased delta modulation at the frontal sensors was the main feature in command following patients. In contrast, higher frequency theta and alpha modulations differentiated controls from both groups of patients. Our findings highlight a key role of spatiotemporally specific delta modulation in supporting cortically mediated behaviour including the ability to follow command. However, patients able to follow commands nevertheless have marked differences in brain activity in comparison with healthy volunteers.

Integrative, multi-omics, analysis of blood samples improves model predictions: applications to cancer

BackgroundCancer genomic studies often include data collected from several omics platforms. Each omics data source contributes to the understanding of the underlying biological process via source specific (“individual”) patterns of variability. At the same time, statistical associations and potential interactions among the different data sources can reveal signals from common biological processes that might not be identified by single source analyses. These common patterns of variability are referred to as “shared” or “joint”. In this work, we show how the use of joint and individual components can lead to better predictive models, and to a deeper understanding of the biological process at hand. We identify joint and individual contributions of DNA methylation, miRNA and mRNA expression collected from blood samples in a lung cancer case–control study nested within the Norwegian Women and Cancer (NOWAC) cohort study, and we use such components to build prediction models for case–control and metastatic status. To assess the quality of predictions, we compare models based on simultaneous, integrative analysis of multi-source omics data to a standard non-integrative analysis of each single omics dataset, and to penalized regression models. Additionally, we apply the proposed approach to a breast cancer dataset from The Cancer Genome Atlas.ResultsOur results show how an integrative analysis that preserves both components of variation is more appropriate than standard multi-omics analyses that are not based on such a distinction. Both joint and individual components are shown to contribute to a better quality of model predictions, and facilitate the interpretation of the underlying biological processes in lung cancer development.ConclusionsIn the presence of multiple omics data sources, we recommend the use of data integration techniques that preserve the joint and individual components across the omics sources. We show how the inclusion of such components increases the quality of model predictions of clinical outcomes.

Multiethnic Prediction of Nicotine Biomarkers and Association With Nicotine Dependence.

IntroductionThe nicotine metabolite ratio and nicotine equivalents are measures of metabolism rate and intake. Genome-wide prediction of these nicotine biomarkers in multiethnic samples will enable tobacco-related biomarker, behavioral, and exposure research in studies without measured biomarkers.Aims and MethodsWe screened genetic variants genome-wide using marginal scans and applied statistical learning algorithms on top-ranked genetic variants, age, ethnicity and sex, and, in additional modeling, cigarettes per day (CPD), (in additional modeling) to build prediction models for the urinary nicotine metabolite ratio (uNMR) and creatinine-standardized total nicotine equivalents (TNE) in 2239 current cigarette smokers in five ethnic groups. We predicted these nicotine biomarkers using model ensembles and evaluated external validity using dependence measures in 1864 treatment-seeking smokers in two ethnic groups.ResultsThe genomic regions with the most selected and included variants for measured biomarkers were chr19q13.2 (uNMR, without and with CPD) and chr15q25.1 and chr10q25.3 (TNE, without and with CPD). We observed ensemble correlations between measured and predicted biomarker values for the uNMR and TNE without (with CPD) of 0.67 (0.68) and 0.65 (0.72) in the training sample. We observed inconsistency in penalized regression models of TNE (with CPD) with fewer variants at chr15q25.1 selected and included. In treatment-seeking smokers, predicted uNMR (without CPD) was significantly associated with CPD and predicted TNE (without CPD) with CPD, time-to-first-cigarette, and Fagerström total score.ConclusionsNicotine metabolites, genome-wide data, and statistical learning approaches developed novel robust predictive models for urinary nicotine biomarkers in multiple ethnic groups. Predicted biomarker associations helped define genetically influenced components of nicotine dependence.ImplicationsWe demonstrate development of robust models and multiethnic prediction of the uNMR and TNE using statistical and machine learning approaches. Variants included in trained models for nicotine biomarkers include top-ranked variants in multiethnic genome-wide studies of smoking behavior, nicotine metabolites, and related disease. Association of the two predicted nicotine biomarkers with Fagerström Test for Nicotine Dependence items supports models of nicotine biomarkers as predictors of physical dependence and nicotine exposure. Predicted nicotine biomarkers may facilitate tobacco-related disease and treatment research in samples with genomic data and limited nicotine metabolite or tobacco exposure data.

Mapping of the World Health Organization's Disability Assessment Schedule 2.0 to disability weights using the Multi-Country Survey Study on Health and Responsiveness.

ObjectivesTo develop and test an internationally applicable mapping function for converting WHODAS‐2.0 scores to disability weights, thereby enabling WHODAS‐2.0 to be used in cost‐utility analyses and sectoral decision‐making.MethodsData from 14 countries were used from the WHO Multi‐Country Survey Study on Health and Responsiveness, administered among nationally representative samples of respondents aged 18+ years who were non‐institutionalized and living in private households. For the combined total of 92,006 respondents, available WHODAS‐2.0 items (for both 36‐item and 12‐item versions) were mapped onto disability weight estimates using a machine learning approach, whereby data were split into separate training and test sets; cross‐validation was used to compare the performance of different regression and penalized regression models. Sensitivity analyses considered different imputation strategies and compared overall model performance with that of country‐specific models.ResultsMapping functions converted WHODAS‐2.0 scores into disability weights; R‐squared values of 0.700–0.754 were obtained for the test data set. Penalized regression models reached comparable performance to standard regression models but with fewer predictors. Imputation had little impact on model performance. Model performance of the generic model on country‐specific test sets was comparable to model performance of country‐specific models.ConclusionsDisability weights can be generated with good accuracy using WHODAS 2.0 scores, including in national settings where health state valuations are not directly available, which signifies the utility of WHODAS as an outcome measure in evaluative studies that express intervention benefits in terms of QALYs gained.

Monitoring sequential structural changes in penalized high-dimensional linear models

In this article, we propose a procedure to monitor the structural changes in the penalized regression model for high-dimensional data sequentially. Our approach utilizes a given historical data set to perform both variable selection and estimation simultaneously. The asymptotic properties of the test statistics are established under the null and alternative hypotheses. The finite sample behavior of the monitoring procedure is investigated with simulation studies. The proposed method is applied to a real data set to illustrate the detection procedure.

A robust variable screening procedure for ultra-high dimensional data.

Variable selection in ultra-high dimensional regression problems has become an important issue. In such situations, penalized regression models may face computational problems and some pre-screening of the variables may be necessary. A number of procedures for such pre-screening has been developed; among them the Sure Independence Screening (SIS) enjoys some popularity. However, SIS is vulnerable to outliers in the data, and in particular in small samples this may lead to faulty inference. In this paper, we develop a new robust screening procedure. We build on the density power divergence (DPD) estimation approach and introduce DPD-SIS and its extension iterative DPD-SIS. We illustrate the behavior of the methods through extensive simulation studies and show that they are superior to both the original SIS and other robust methods when there are outliers in the data. Finally, we illustrate its use in a study on regulation of lipid metabolism.

Risk factors for faecal carriage of multidrug-resistant Escherichia coli in a college community: a penalised regression model

Bacterial antimicrobial resistance is a serious global public-health threat. Intestinal commensal drug-resistant bacteria have been suggested as an important reservoir of antimicrobial resistance genes (ARGs), which may be acquired via food. We aimed to identify risk factors associated with faecal carriage of drug-resistant commensal Escherichia coli among healthy adults focused on their dietary habits. We conducted a cross-sectional study targeting healthy adult volunteers in a college community. Faecal samples and questionnaires were obtained from 113 volunteers. We conducted backward elimination logistic regression and least absolute shrinkage and selection operator (LASSO) methods to identify risk factors. We analysed responses from 81 of 113 volunteers who completed the questionnaire. The logistic regression and LASSO methods identified red meat consumption to be associated with an increased risk [OR=6.13 (95% CI 1.83-24.2) and 1.82, respectively] and fish consumption with a reduced risk [OR=0.27 (95% CI 0.08-0.85) and 0.81] for carriage of multidrug-resistant (MDR) E. coli, adjusted for biological sex, employment status, frequently used supermarket and previous travel. Dietary habits are associated with risk of faecal carriage of MDR E. coli. This study supports the growing evidence that food may be an important source of ARGs present in human commensal E. coli.

Abstract 041: Cardiovascular Disease, Hypertension, Diabetes And Cystatin C Jointly Predict Covid-19 Mortality Alongside Age, Male Sex And Black Ethnicity

Introduction: Studies of risk factors for severe/fatal COVID-19 to date may not have identified the optimal set of informative predictors. Hypothesis: Use of penalized regression with stability analysis may identify new, sparse sets of risk factors jointly associated with COVID-19 mortality. Methods: We investigated demographic, social, lifestyle, biological (lipids, cystatin C, vitamin D), medical (comorbidities, medications) and air pollution data from UK Biobank (N=473,574) in relation to linked COVID-19 mortality, and compared with non-COVID-19 mortality. We used penalized regression models (LASSO) with stability analysis (80% selection threshold from 1,000 models with 80% subsampling) to identify a sparse set of variables associated with COVID-19 mortality. Results: Among 43 variables considered by LASSO stability selection, cardiovascular disease, hypertension, diabetes, cystatin C, age, male sex and Black ethnicity were jointly predictive of COVID-19 mortality risk at 80% selection threshold (Figure). Of these, Black ethnicity and hypertension contributed to COVID-19 but not non-COVID-19 mortality. Conclusions: Use of LASSO stability selection identified a sparse set of predictors for COVID-19 mortality including cardiovascular disease, hypertension, diabetes and cystatin C, a marker of renal function that has also been implicated in atherogenesis and inflammation. These results indicate the importance of cardiometabolic comorbidities as predisposing factors for COVID-19 mortality. Hypertension was differentially highly selected for risk of COVID-19 mortality, suggesting the need for continued vigilance with good blood pressure control during the pandemic.

Machine-learning based prediction of Cushing\u2019s syndrome in dogs attending UK primary-care veterinary practice

Cushing’s syndrome is an endocrine disease in dogs that negatively impacts upon the quality-of-life of affected animals. Cushing’s syndrome can be a challenging diagnosis to confirm, therefore new methods to aid diagnosis are warranted. Four machine-learning algorithms were applied to predict a future diagnosis of Cushing's syndrome, using structured clinical data from the VetCompass programme in the UK. Dogs suspected of having Cushing's syndrome were included in the analysis and classified based on their final reported diagnosis within their clinical records. Demographic and clinical features available at the point of first suspicion by the attending veterinarian were included within the models. The machine-learning methods were able to classify the recorded Cushing’s syndrome diagnoses, with good predictive performance. The LASSO penalised regression model indicated the best overall performance when applied to the test set with an AUROC = 0.85 (95% CI 0.80–0.89), sensitivity = 0.71, specificity = 0.82, PPV = 0.75 and NPV = 0.78. The findings of our study indicate that machine-learning methods could predict the future diagnosis of a practicing veterinarian. New approaches using these methods could support clinical decision-making and contribute to improved diagnosis of Cushing’s syndrome in dogs.

URINARY METABOLIC PHENOTYPE OF BLOOD PRESSURE

Objective: Metabolic phenotyping (metabolomics) captures systems-level information on metabolic processes by simultaneously measuring hundreds of metabolites using spectroscopic techniques. Concentrations of these metabolites are affected by genetic (host, microbiome), environmental and dietary factors and may provide insights into biochemical pathways underlying raised blood pressure (BP) in populations. Design and method: Two separate, timed 24hr urine specimens were obtained from 2,031 women and men, aged 40–59, from 8 USA population samples in the INTERMAP Study. Proton Nuclear Magnetic Resonance (1H NMR) was used to characterize a urinary metabolic signature; this was unaffected by diurnal variability and sampling time as it captures end-products of metabolism over a 24hr period. Demographic, population, medical, lifestyle and anthropometric factors were accounted for in regression models to define a urinary metabolic phenotype associated with BP. Results: 29 structurally identified urinary metabolites covaried with systolic BP (SBP), after adjustment for demographic variables, and 18 metabolites with diastolic BP (DBP), with 16 metabolites overlapping between SBP and DBP. These included metabolites related to energy metabolism, renal function, diet and gut microbiota. After adjustment for medical and lifestyle covariates, 22/14 metabolites remained associated with SBP/DBP. Joint covariate-metabolite penalized regression models identified Body Mass Index, age and family history as most important contributors, with 14 metabolites, including gut microbial co-metabolites, also included in the model. Metabolites were mapped in a symbiotic metabolic reaction network, that includes reactions mediated by 3,344 commensal gut microbial species, to highlight affected pathways (Figure). Significant single nucleotide polymorphisms (SNPs) from genome-wide association studies on cardiometabolic risk factors were mapped to genes in this network. This revealed multiple subnetworks of gene-metabolite pairs related to BP and related cardiometabolic factors and includes 54 SNPs directly related to reactions in the network. These 54 SNPs were then used as instrumental variables to test for possible causative metabolite-BP associations in an external cohort (Airwave Study). Conclusions: These results highlight the complex interplay between human genes, the microbiome and metabolites associated with BP and other cardiometabolic risk factors.

COVID-19 mortality in the UK Biobank cohort: revisiting and evaluating risk factors

Most studies of severe/fatal COVID-19 risk have used routine/hospitalisation data without detailed pre-morbid characterisation. Using the community-based UK Biobank cohort, we investigate risk factors for COVID-19 mortality in comparison with non-COVID-19 mortality. We investigated demographic, social (education, income, housing, employment), lifestyle (smoking, drinking, body mass index), biological (lipids, cystatin C, vitamin D), medical (comorbidities, medications) and environmental (air pollution) data from UK Biobank (N = 473,550) in relation to 459 COVID-19 and 2626 non-COVID-19 deaths to 21 September 2020. We used univariate, multivariable and penalised regression models. Age (OR = 2.76 [2.18–3.49] per S.D. [8.1 years], p = 2.6 × 10–17), male sex (OR = 1.47 [1.26–1.73], p = 1.3 × 10–6) and Black versus White ethnicity (OR = 1.21 [1.12–1.29], p = 3.0 × 10–7) were independently associated with and jointly explanatory of (area under receiver operating characteristic curve, AUC = 0.79) increased risk of COVID-19 mortality. In multivariable regression, alongside demographic covariates, being a healthcare worker, current smoker, having cardiovascular disease, hypertension, diabetes, autoimmune disease, and oral steroid use at enrolment were independently associated with COVID-19 mortality. Penalised regression models selected income, cardiovascular disease, hypertension, diabetes, cystatin C, and oral steroid use as jointly contributing to COVID-19 mortality risk; Black ethnicity, hypertension and oral steroid use contributed to COVID-19 but not non-COVID-19 mortality. Age, male sex and Black ethnicity, as well as comorbidities and oral steroid use at enrolment were associated with increased risk of COVID-19 death. Our results suggest that previously reported associations of COVID-19 mortality with body mass index, low vitamin D, air pollutants, renin–angiotensin–aldosterone system inhibitors may be explained by the aforementioned factors.

A fuzzy penalized regression model with variable selection

In the classical multiple regression modeling, there might be some insignificant input variables. These variables can be eliminated by automatic selectors, known as penalized methods. We propose a penalized estimation method for the coefficients of a linear regression model for studying the dependence of a LR fuzzy response (output) variable on a set of crisp explanatory (input) variables. To show the performances of the proposed model a simulation study was utilized under three scenarios of multicollinear and sparse data. The model demonstrates better performances in comparison to another three models on the basis of specified goodness of fit measures, under three variants of the penalty function. Evaluation of the method has been conducted on real data. The results demonstrate superior performances in terms of the goodness of fit measures in comparison to the other models. To take into account the imprecision due to the lack of knowledge about the data generation process, in the applications to real data bootstrap-t confidence intervals were also utilized.

Prospective Identification of Elevated Circulating CDCP1 in Patients Years before Onset of Lung Cancer.

Increasing evidence points to a role for inflammation in lung carcinogenesis. A small number of circulating inflammatory proteins have been identified as showing elevated levels prior to lung cancer diagnosis, indicating the potential for prospective circulating protein concentration as a marker of early carcinogenesis. To identify novel markers of lung cancer risk, we measured a panel of 92 circulating inflammatory proteins in 648 prediagnostic blood samples from two prospective cohorts in Italy and Norway (women only). To preserve the comparability of results and protect against confounding factors, the main statistical analyses were conducted in women from both studies, with replication sought in men (Italian participants). Univariate and penalized regression models revealed for the first time higher blood levels of CDCP1 protein in cases that went on to develop lung cancer compared with controls, irrespective of time to diagnosis, smoking habits, and gender. This association was validated in an additional 450 samples. Associations were stronger for future cases of adenocarcinoma where CDCP1 showed better explanatory performance. Integrative analyses combining gene expression and protein levels of CDCP1 measured in the same individuals suggested a link between CDCP1 and the expression of transcripts of LRRN3 and SEM1. Enrichment analyses indicated a potential role for CDCP1 in pathways related to cell adhesion and mobility, such as the WNT/β-catenin pathway. Overall, this study identifies lung cancer-related dysregulation of CDCP1 expression years before diagnosis. SIGNIFICANCE: Prospective proteomics analyses reveal an association between increased levels of circulating CDCP1 and lung carcinogenesis irrespective of smoking and years before diagnosis, and integrating gene expression indicates potential underlying mechanisms.See related commentary by Itzstein et al., p. 3441.

A non-convex regularization approach for stable estimation of loss development factors

In this article, we apply non-convex regularization methods in order to obtain stable estimation of loss development factors in insurance claims reserving. Among the non-convex regularization methods, we focus on the use of the log-adjusted absolute deviation (LAAD) penalty and provide discussion on optimization of LAAD penalized regression model, which we prove to converge with a coordinate descent algorithm under mild conditions. This has the advantage of obtaining a consistent estimator for the regression coefficients while allowing for the variable selection, which is linked to the stable estimation of loss development factors. We calibrate our proposed model using a multi-line insurance dataset from a property and casualty insurer where we observed reported aggregate loss along accident years and development periods. When compared to other regression models, our LAAD penalized regression model provides very promising results.