Double Cross-validation Procedure Research Articles

Comprehensive characterisation of Culicoides clastrieri and C. festivipennis (Diptera: Ceratopogonidae) according to morphological and morphometric characters using a multivariate approach and DNA barcode

Biting midges are widespread around the world and play an essential role in the epidemiology of over 100 veterinary and medical diseases. For taxonomists, it is difficult to correctly identify species because of affinities among cryptic species and species complexes. In this study, species boundaries were examined for C. clastrieri and C. festivipennis and compared with six other Culicoides species. The classifiers are partial least squares discriminant analysis (PLS-DA) and sparse partial least squares discriminant analysis (sPLS-DA).The performance of the proposed method was evaluated using four models: (i) geometric morphometrics applied to wings; (ii) morphological wing characters, (iii) "Full wing" (landmarks and morphological characters) and (iv) "Full model" (morphological characters—wing, head, abdomen, legs—and wing landmarks). Double cross-validation procedures were used to validate the predictive ability of PLS-DA and sPLS-DA models. The AUC (area under the ROC curve) and the balanced error rate showed that the sPLS-DA model performs better than the PLS-DA model. Our final sPLS-DA analysis on the full wing and full model, with nine and seven components respectively, managed to perfectly classify our specimens. The C. clastrieri and C. festivipennis sequences, containing both COI and 28S genes, revealed our markers’ weak discrimination power, with an intraspecific and interspecific divergence of 0.4% and 0.1% respectively. Moreover, C. clastrieri and C. festivipennis are grouped in the same clade. The morphology and wing patterns of C. clastrieri and C. festivipennis can be used to clearly distinguish them. Our study confirms C. clastrieri and C. festivipennis as two distinct species. Our results show that caution should be applied when relying solely on DNA barcodes for species identification or discovery.

Towards Deep Neural Network Models for the Prediction of the Blood-Brain Barrier Permeability for Diverse Organic Compounds.

Permeation through the blood–brain barrier (BBB) is among the most important processes controlling the pharmacokinetic properties of drugs and other bioactive compounds. Using the fragmental (substructural) descriptors representing the occurrence number of various substructures, as well as the artificial neural network approach and the double cross-validation procedure, we have developed a predictive in silico LogBB model based on an extensive and verified dataset (529 compounds), which is applicable to diverse drugs and drug-like compounds. The model has good predictivity parameters (, ) that are similar to or better than those of the most reliable models available in the literature. Larger datasets, and perhaps more sophisticated network architectures, are required to realize the full potential of deep neural networks. The analysis of fragment contributions reveals patterns of influence consistent with the known concepts of structural characteristics that affect the BBB permeability of organic compounds. The external validation of the model confirms good agreement between the predicted and experimental LogBB values for most of the compounds. The model enables the evaluation and optimization of the BBB permeability of potential neuroactive agents and other drug compounds.

Resting-state prefrontal EEG biomarkers in correlation with MMSE scores in elderly individuals

We investigated whether cognitive decline could be explained by resting-state electroencephalography (EEG) biomarkers measured in prefrontal regions that reflect the slowing of intrinsic EEG oscillations. In an aged population dwelling in a rural community (total = 496, males = 165, females = 331), we estimated the global cognitive decline using the Mini-Mental State Examination (MMSE) and measured resting-state EEG parameters at the prefrontal regions of Fp1 and Fp2 in an eyes-closed state. Using a tertile split method, the subjects were classified as T3 (MMSE 28–30, N = 162), T2 (MMSE 25–27, N = 179), or T1 (MMSE ≤ 24, N = 155). The EEG slowing biomarkers of the median frequency, peak frequency and alpha-to-theta ratio decreased as the MMSE scores decreased from T2 to T1 for both sexes (−5.19 ≤ t-value ≤ −3.41 for males and −7.24 ≤ t-value ≤ −4.43 for females) after adjusting for age and education level. Using a double cross-validation procedure, we developed a prediction model for the MMSE scores using the EEG slowing biomarkers and demographic covariates of sex, age and education level. The maximum intraclass correlation coefficient between the MMSE scores and model-predicted values was 0.757 with RMSE = 2.685. The resting-state EEG biomarkers showed significant changes in people with early cognitive decline and correlated well with the MMSE scores. Resting-state EEG slowing measured in the prefrontal regions may be useful for the screening and follow-up of global cognitive decline in elderly individuals.

Inflammatory signatures in older persons with physical frailty and sarcopenia: The frailty "cytokinome" at its core.

The construct of physical frailty and sarcopenia (PF&S) identifies an age-related pre-disability condition defined by reduced physical performance and low muscle mass. Whether PF&S is characterized by perturbations of the cytokine network is presently unclear. Furthermore, the existence of gender-specific inflammatory profiles of PF&S is unknown. This study was designed to explore the association between a large panel of inflammatory biomolecules and PF&S in older adults through a multivariate statistical approach. Gender-specific inflammatory patterns were also explored. One-hundred community-dwellers aged 70 years and older with PF&S and 100 non-sarcopenic, non-frail controls (nonPF&S) were enrolled. A panel of 30 circulating inflammatory biomarkers was assayed. Partial least squares discriminant analysis (PLS-DA) was employed to explore the relationship between inflammatory molecules and PF&S. Separate PLS-DA models were built for the whole sample and the two genders. Double cross-validation procedures were used to validate the predictive ability of PLS-DA models. The optimal complexity of the PLS-DA model built on the whole sample was found to be four latent variables. The proportion of correct classification was 75.6 ± 1.3% (82.3 ± 1.6% for enrollees with PF&S and 68.7 ± 2.5% for nonPF&S controls). The inflammatory profile of people with PF&S was defined by higher levels of P-selectin, C-reactive protein (CRP), and interferon γ-induced protein 10. NonPF&S participants were characterized by higher levels of myeloperoxidase (MPO), interleukin (IL) 8, monocyte chemoattractant protein 1 (MCP-1), macrophage inflammatory protein 1-α, platelet-derived growth factor (PDGF) BB. Gender-specific PLS-DA allowed identifying a "core" inflammatory signature of PF&S, composed by higher levels of CRP, and lower concentrations of MPO, IL8, MCP-1, and PDGF-BB, with peculiar patterns of relationships for men and women. A core inflammatory profile was identified in people with PF&S with a gender-specific signature. The dissection of the PF&S "cytokinome" will provide novel insights into the role played by inflammation in the disabling cascade and allow designing personalized treatment strategies.

Variable selection and validation in multivariate modelling.

MotivationValidation of variable selection and predictive performance is crucial in construction of robust multivariate models that generalize well, minimize overfitting and facilitate interpretation of results. Inappropriate variable selection leads instead to selection bias, thereby increasing the risk of model overfitting and false positive discoveries. Although several algorithms exist to identify a minimal set of most informative variables (i.e. the minimal-optimal problem), few can select all variables related to the research question (i.e. the all-relevant problem). Robust algorithms combining identification of both minimal-optimal and all-relevant variables with proper cross-validation are urgently needed.ResultsWe developed the MUVR algorithm to improve predictive performance and minimize overfitting and false positives in multivariate analysis. In the MUVR algorithm, minimal variable selection is achieved by performing recursive variable elimination in a repeated double cross-validation (rdCV) procedure. The algorithm supports partial least squares and random forest modelling, and simultaneously identifies minimal-optimal and all-relevant variable sets for regression, classification and multilevel analyses. Using three authentic omics datasets, MUVR yielded parsimonious models with minimal overfitting and improved model performance compared with state-of-the-art rdCV. Moreover, MUVR showed advantages over other variable selection algorithms, i.e. Boruta and VSURF, including simultaneous variable selection and validation scheme and wider applicability.Availability and implementationAlgorithms, data, scripts and tutorial are open source and available as an R package (‘MUVR’) at https://gitlab.com/CarlBrunius/MUVR.git.Supplementary information Supplementary data are available at Bioinformatics online.

Toward Spectral Library-Free Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry Bacterial Identification.

Bacterial identification is of great importance in clinical diagnosis, environmental monitoring, and food safety control. Among various strategies, matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) has drawn significant interest and has been clinically used. Nevertheless, current bioinformatics solutions use spectral libraries for the identification of bacterial strains. Spectral library generation requires acquisition of MALDI-TOF spectra from monoculture bacterial colonies, which is time-consuming and not possible for many species and strains. We propose a strategy for bacterial typing by MALDI-TOF using protein sequences from public database, that is, UniProt. Ten genes were identified to encode proteins most often observed by MALD-TOF from bacteria through 500 times repeated a 10-fold double cross-validation procedure, using 403 MALDI-TOF spectra corresponding to 14 genera, 81 species, and 403 strains, and the protein sequences of 1276 species in UniProt. The 10 genes were then used to annotate peaks on MALDI-TOF spectra of bacteria for bacterial identification. With the approach, bacteria can be identified at the genus level by searching against a database containing the protein sequences of 42 genera of bacteria from UniProt. Our approach identified 84.1% of the 403 spectra correctly at the genus level. Source code of the algorithm is available at https://github.com/dipcarbon/BacteriaMSLF.

NMR-based metabolomic approach to study urine samples of chronic inflammatory rheumatic disease patients.

The nuclear magnetic resonance (NMR)-based metabolomic approach was used as analytical methodology to study the urine samples of chronic inflammatory rheumatic disease (CIRD) patients. The urine samples of CIRD patients were compared to the ones of both healthy subjects and patients with multiple sclerosis (MS), another immuno-mediated disease. Urine samples collected from 39 CIRD patients, 25 healthy subjects, and 26 MS patients were analyzed using 1H NMR spectroscopy, and the NMR spectra were examined using partial least squares-discriminant analysis (PLS-DA). PLS-DA models were validated by a double cross-validation procedure and randomization tests. Clear discriminations between CIRD patients and healthy controls (average diagnostic accuracy 83.5 ± 1.9%) as well as between CIRD patients and MS patients (diagnostic accuracy 81.1 ± 1.9%) were obtained. Leucine, alanine, 3-hydroxyisobutyric acid, hippuric acid, citric acid, 3-hydroxyisovaleric acid, and creatinine contributed to the discrimination; all of them being in a lower concentration in CIRD patients as compared to controls or to MS patients. The application of NMR metabolomics to study these still poorly understood diseases can be useful to better clarify the pathologic mechanisms; moreover, as a holistic approach, it allowed the detection of, by means of anomalous metabolic traits, the presence of other pathologies or pharmaceutical treatments not directly connected to CIRDs, giving comprehensive information on the general health state of individuals. Graphical abstract NMR-based metabolomic approach as a tool to study urine samples in CIRD patients with respect to MS patients and healthy controls.

TIPOVI LIČNOSTI U OKVIRU MODELA VELIKIH PET: POREĐENJE REŠENJA DOBIJENIH KLASTER ANALIZOM I ANALIZOM LATENTNIH PROFILA

Osnovni cilj ovog istraživanja bio je da se utvrdi broj i sadržaj tipova ličnosti zasnovanih na skorovima ispitanika na dimenzijama modela Velikih pet. Za utvrđivanje tipova ličnosti korišćene su klaster analiza i analiza latentnih profila, te je drugi cilj istraživanja bilo ispitivanje kongruencije rešenja ekstrahovanih ovim metodama. Uzorak se sastojao od 842 ispitanika oba pola i starosti između 18 i 68 godina. Primenjen je upitnik BFI koji sadrži 44 stavke sa petostepenom skalom Likertovog tipa. Upitnik je pokazao zadovoljavajuća metrijska svojstva. Primenom dvoetapne kros-validacione procedure ekstrahovana su tri klastera. U analizi latentnih profila, rešenje sa tri klase se, takođe, pokazalo optimalnim. U oba slučaja, tipovi su imenovani kao „suzdržani“,„rezilijentni“ i „neadaptirani“. Suzdržani postižu prosečne skorove na svim dimenzijama, rezilijentni imaju niske skorove na skali Neuroticizam, a visoke na svim drugim dimenzijama, a ispitanici iz klastera neadaptiranih postižu visoke skorove na skali Neuroticizam, a niske na svim drugim dimenzijama. Vrednost Cohenovog kapa koeficijenta (κ = .70) ukazuje na visoku kongruenciju dva dobijena rešenja. Profil rezilijentnih, koji se izdvaja u okviru dominantne ARC (Asendorpf-Robins-Caspi) tipologije, dobijen je i u ovom istraživanju, što upućuje na njegovu kros-kulturalnu stabilnost. S druge strane, suzdržani i neadaptirani profili nisu u potpunosti podudarni s profilima hiperkontrolisanih i hipokontrolisanih iz ARC tipologije, iako je profil neadaptiranih, najčešće pod nazivom „nepoželjni“ izdvojen i u pojedinim drugim studijama. Rezultati ukazuju na značajnu, ali ne i potpunu, podudarnost sadržaja solucija ekstrahovanih dvema metodama, kao i na izvesne razlike između tipova ličnosti ekstrahovanih u našoj populaciji i rešenja identifikovanih u drugim kulturama.

A comparative investigation of modern feature selection and classification approaches for the analysis of mass spectrometry data

Many analytical approaches such as mass spectrometry generate large amounts of data (input variables) per sample analysed, and not all of these variables are important or related to the target output of interest. The selection of a smaller number of variables prior to sample classification is a widespread task in many research studies, where attempts are made to seek the lowest possible set of variables that are still able to achieve a high level of prediction accuracy; in other words, there is a need to generate the most parsimonious solution when the number of input variables is huge but the number of samples/objects are smaller. Here, we compare several different variable selection approaches in order to ascertain which of these are ideally suited to achieve this goal. All variable selection approaches were applied to the analysis of a common set of metabolomics data generated by Curie-point pyrolysis mass spectrometry (Py-MS), where the goal of the study was to classify the Gram-positive bacteria Bacillus. These approaches include stepwise forward variable selection, used for linear discriminant analysis (LDA); variable importance for projection (VIP) coefficient, employed in partial least squares-discriminant analysis (PLS-DA); support vector machines-recursive feature elimination (SVM-RFE); as well as the mean decrease in accuracy and mean decrease in Gini, provided by random forests (RF). Finally, a double cross-validation procedure was applied to minimize the consequence of overfitting. The results revealed that RF with its variable selection techniques and SVM combined with SVM-RFE as a variable selection method, displayed the best results in comparison to other approaches.

Found in transcription: Accurate Parkinson's disease classification in peripheral blood

A blood-based test for the early detection of Parkinson's disease (PD) would be an important diagnostic tool and useful for patient selection when developing novel drugs or treatments for the disease. Here, we aimed to identify potential biomarkers associated with PD. We applied gene expression profiling to the study of peripheral blood from 75 healthy control subjects and 79 PD patients at different stages of the disease. Healthy control subjects were matched for age and gender with PD subjects, and the diagnosis of patients was based on clinical evaluation by specialists in movement disorders. RNA was extracted from the blood samples and the gene expressions were measured using the Illumina HumanHT-12 v4.0 Expression BeadChip. Our results support previous studies that gene expression in blood may be instrumental in the search for molecular biomarkers for PD. Single cross-validation results show that PD can be correctly classified from healthy controls with an agreement of 88% to clinical diagnosis. De novo PD patients are classified with a sensitivity of 87%, which is close to what was achieved for the patients having a confirmed PD diagnosis with disease duration <5 and >5 years (93% and 88%). A double cross-validation procedure showed that using a selected set of around 650 informative genes, similar results are achieved. Functional analysis of the selected genes showed genes significantly associated to mitochondrial dysfunction, protein ubiquitination, gene expression and cell death. PD affects gene expression in blood, suggesting the potential for the development of a blood-based gene expression test.

Double-check: validation of diagnostic statistics for PLS-DA models in metabolomics studies

Partial Least Squares-Discriminant Analysis (PLS-DA) is a PLS regression method with a special binary ‘dummy’ y-variable and it is commonly used for classification purposes and biomarker selection in metabolomics studies. Several statistical approaches are currently in use to validate outcomes of PLS-DA analyses e.g. double cross validation procedures or permutation testing. However, there is a great inconsistency in the optimization and the assessment of performance of PLS-DA models due to many different diagnostic statistics currently employed in metabolomics data analyses. In this paper, properties of four diagnostic statistics of PLS-DA, namely the number of misclassifications (NMC), the Area Under the Receiver Operating Characteristic (AUROC), Q 2 and Discriminant Q 2 (DQ 2) are discussed. All four diagnostic statistics are used in the optimization and the performance assessment of PLS-DA models of three different-size metabolomics data sets obtained with two different types of analytical platforms and with different levels of known differences between two groups: control and case groups. Statistical significance of obtained PLS-DA models was evaluated with permutation testing. PLS-DA models obtained with NMC and AUROC are more powerful in detecting very small differences between groups than models obtained with Q 2 and Discriminant Q 2 (DQ 2). Reproducibility of obtained PLS-DA models outcomes, models complexity and permutation test distributions are also investigated to explain this phenomenon. DQ 2 and Q 2 (in contrary to NMC and AUROC) prefer PLS-DA models with lower complexity and require higher number of permutation tests and submodels to accurately estimate statistical significance of the model performance. NMC and AUROC seem more efficient and more reliable diagnostic statistics and should be recommended in two group discrimination metabolomic studies.Electronic supplementary materialThe online version of this article (doi:10.1007/s11306-011-0330-3) contains supplementary material, which is available to authorized users.

Application of Kinase Activity Profiles to Predict Upcoming TKI Resistance In CML-Patients.

AbstractAbstract 3425 Background:Major breakthroughs in CML research have revolutionized therapy by the introduction of a specific inhibitor of BCR-ABL activity; first Imatinib and now even more potent TKIs like Nilotinib and Dasatinib are available. Imatinib can induce complete hematologic responses in virtually all patients and complete cytogenetic responses in almost 90% of patients and progression free survival in 84% of patients. Patients that reach a more than 1000-fold reduction of the BCR-ABL transcript level (major molecular response) have durable responses with not a single event of progression. Unfortunately, not all patients attain such a favourable response. Primary and secondary resistance develops in around 20% of cases.Predicting the development this resistance would be of great clinical value. Aim:To establish a classifier based on peptide phosphorylation patterns as fingerprints using samples of peripheral blood cells or bone marrow of CML patients undergoing Imatinib treatment in order to enable prediction of emerging resistance against Imatinib. Methods:Stored (“snapfrozen”) mononuclear cell fractions, isolated by Ficoll-Paque centrifugation of bone marrow or peripheral blood of CML patients under continuous Imatinib treatment and in various stages of their disease were lysed in M-PER buffer supplemented with phosphatase and protease inhibitors. Kinase activity profiles of these lysates were generated with standard Tyrosine Kinase PamChip® Arrays that contain 144 peptides as kinase substrates on their porous surface. Peptide phosphorylation was followed through binding of a fluorescently labelled anti-phosphotyrosine antibody during flow-through cycling of the lysates. Activity profiles were generated with PamGene's BioNavigator software. The resulting data were analysed using the R-based package CMA (“Classification for MicroArrays”) that enables the survey and evaluation of most usual classification methods with double cross-validation procedures. Results:A distinct classifier for Imatinib response prediction could be derived for the bone marrow samples, collected at various intervals after diagnosis (18 Imatinib sensitive versus 19 Imatinib resistant patients, of which the samples were collected 3 months to 1.5 year before resistance emergence). Of the twenty-one classification methods that were studied the SupportVectorMachine (SVM) algorithm resulted in the smallest error rate: this was 16% with a standard error of 0.062 and a sensitivity of 84% and specificity of 82% respectively, meaning 3 misclassified sensitive patients and 3 misclassified resistant patients. This classifier has to be validated in a blinded, independent test set. Conclusions:We demonstrated that with this method it is possible to predict Imatinib resistance. Differences in phosphorylation patterns as detected using PamGene's peptide microarray technology with the aid of multivariate statistical analysis suggest the presence of an ongoing process in CML-patients destined in due time to a relapse in spite of continuous Imatinib treatment. We regard these initial class prediction results to be a basis for further development of this kinase activity based test for Imatinib response prediction in CML patients already at the time of diagnosis.Perspectives: Our approach of using kinase activity profiling for the prediction of Imatinib resistance, could equally well be applied to predict response to various other kinase inhibitors. These response predictions could be combined in the same test by adding these inhibitors in vitro and might direct the optimal drug selection at the time point of diagnosis. Disclosures:Boender:PamGene International BV: Employment. Ruijtenbeek:PamGene International BV: Employment. van den Berg:PamGene Inetrnational BV: Employment. de Wijn:PamGene International BV: Employment.

Verification of the geographical origin of European butters using PTR-MS

In the present study, proton transfer reaction-mass spectrometry (PTR-MS) in combination with partial least square-discriminant analysis (PLS-DA) was evaluated as a method for the prediction of the origin of European butters. Eighty-three commercial butters from three European regions were subjected to headspace analysis using PTR-MS. Data were collected for the mass range m/z 20–150 using a dwell time of 0.2 s mass −1, resulting in a cycle time just under 30 s. The log transformed headspace concentrations of the masses were subjected to PLS-DA in order to estimate classification models for the butter samples. One model predicted the region of origin; a second set of models predicted dichotomously whether or not a butter originated from a particular EU country. The performance of each model was evaluated by means of a 10-fold double cross validation procedure. For 76% of the butters the region of origin was predicted correctly in the cross validation. The success rate of the countries, averaged over all dichotomous models, was 88% but large differences between countries were observed. Additional work is required to study the underlying factors that determine the geographical differences in butter volatile compositions.

Calculation of the stability of β-cyclodextrin complexes of organic compounds using the QSPR approach

The QSPR model for predicting the Gibbs free energy of formation of complexes of different organic compounds with β-cyclodextrin has been constructed by the multiple linear regression method with the use of the double cross-validation procedure. The model has good predictive power.

Customised birthweight: Coefficients for an Australian population and validation of the model

Published birthweight references in Australia do not fully take into account constitutional factors that influence birthweight and therefore may not provide an accurate reference to identify the infant with abnormal growth. Furthermore, studies in other regions that have derived adjusted (customised) birthweight references have applied untested assumptions in the statistical modelling. To validate the customised birthweight model and to produce a reference set of coefficients for estimating a customised birthweight that may be useful for maternity care in Australia and for future research. De-identified data were extracted from the clinical database for all births at the Mater Mother's Hospital, Brisbane, Australia, between January 1997 and June 2005. Births with missing data for the variables under study were excluded. In addition the following were excluded: multiple pregnancies, births less than 37 completed week's gestation, stillbirths, and major congenital abnormalities. Multivariate analysis was undertaken. A double cross-validation procedure was used to validate the model. The study of 42,206 births demonstrated that, for statistical purposes, birthweight is normally distributed. Coefficients for the derivation of customised birthweight in an Australian population were developed and the statistical model is demonstrably robust. This study provides empirical data as to the robustness of the model to determine customised birthweight. Further research is required to define where normal physiology ends and pathology begins, and which segments of the population should be included in the construction of a customised birthweight standard.

Typification of Cider Brandy on the Basis of Cider Used in Its Manufacture

A study of typification of cider brandies on the basis of the origin of the raw material used in their manufacture was conducted using chemometric techniques (principal component analysis, linear discriminant analysis, and Bayesian analysis) together with their composition in volatile compounds, as analyzed by gas chromatography with flame ionization to detect the major volatiles and by mass spectrometric to detect the minor ones. Significant principal components computed by a double cross-validation procedure allowed the structure of the database to be visualized as a function of the raw material, that is, cider made from fresh apple juice versus cider made from apple juice concentrate. Feasible and robust discriminant rules were computed and validated by a cross-validation procedure that allowed the authors to classify fresh and concentrate cider brandies, obtaining classification hits of >92%. The most discriminating variables for typifying cider brandies according to their raw material were 1-butanol and ethyl hexanoate.

Characterization of Cider Brandy on the Basis of Aging Time

ABSTRACT: To characterize cider brandies on the basis of age, chemometric techniques (principal component analysis, linear discriminant analysis, and Bayesean analysis) together with contents of volatile, furanic, and phenolic compounds were used. Significant principal components computed by a double cross‐validation procedure allowed us to visualize the structure of the database as a function of the aging time to which the cider brandies had been subjected. Feasible and robust discriminant rules were computed and validated by a cross‐validation procedure that allowed suitable classification of fresh and aged brandies, obtaining classification hits &gt;; 90%. The most discriminant variables for characterizing cider brandies according to their aging level were as follows: ethyl caprylate &gt; ethyl isovalerate &gt; 1‐propanol &gt; hexyl acetate.

Near-Global Sea Surface Temperature Anomalies as Predictors of Australian Seasonal Rainfall

An operational system for the prediction of Australian seasonal rainfall variations using sea surface temperature anomaly (SSTA) patterns over the Indian and Pacific Oceans is described. The SSTA patterns are represented by rotated principal components, with individual monthly values at 1- and 3-month lead times used as predictors; for example, November and January SSTAs are used to forecast March‐May seasonal rainfall. The historical seasonal rainfall is also represented by rotated principal components of a gridded 1 8 rainfall dataset, with the principal component loadings used as weights to project the forecasts back to the original 1 8 grid points. Forecasts of seasonal rainfall in two (above/below median) or three categories (terciles) are produced using linear discriminant analysis. Hindcast skill, measured by the linear error in probability space (LEPS) skill score has been assessed using cross validation. Experiments were also performed using a double or nested crossvalidation procedure to select the best model or combination of predictors. The model chosen for operational seasonal forecasts uses the first two rotated SSTA components lagged by 1 and 3 months as predictors for every season and location, to maintain continuity of forecast probabilities between the overlapping 3-month seasons. Current values of the principal component amplitudes are calculated by projecting either the Bureau of Meteorology’s or the National Centers for Environmental Prediction’s SST analysis onto the set of SST principal components. The hindcasts and experimental real-time forecasts over the 5-yr period from January‐March 1994 to December‐February 1998/99 indicate improved skill over parts of southern Australia during the autumn period using the SST-based schemes when compared with forecasts using the Southern Oscillation index alone.

Use of Double Cross-Validation and Bootstrap Methods to Estimate Replicability of Results of Multiple Regression

Researchers have all too often ignored replicability of results because they overly rely on significance testing. This is a misinformed view because statistical significance does not evaluate the importance or replicability of a result. This paper focuses on two methods of assessing replicability of results, double cross-validation and bootstrap procedures. Selected variables from Hughes, Cavell, and Grossman's (1997) data set of 112 cases are used to illustrate all these techniques as applied to the interpretation of multiple regression results. One statistical computer package SPSS is used for the double cross-validation procedure, and a 1987 microcomputer program package of Lunneborg is used to demonstrate the bootstrap procedure.

Binge eating in obesity: Toward the specification of predictors

Obese binge eaters evidence greater disturbance than do obese nonbingers on various eating and weight-related characteristics and on many psychopathological dimensions as well. The present study attempted to specify the nature of risk factors for binging further by regressing factors from the Eating Disorders Inventory (EDI) and the Three-Factor Eating Questionnaire onto a measure of binge eating. Five variables were shown to reliably predict binge eating in a double cross-validation procedure. Of these 5 variables, four assessed eating and weight-related characteristics rather than psychopathology. The possible relevance of these predictor variables for understanding the causes of binge eating was considered.