Variable selection for case-cohort studies with failure time outcome

I evaluated the predictive ability of statistical models obtained by applying seven methods of variable selection to 12 ecological and environmental data sets. Cross-validation, involving repeated splits of each data set into training and validation subsets, was used to obtain honest estimates of predictive ability that could be fairly compared among methods. There was surprisingly little difference in predictive ability among five methods based on multiple linear regression. Stepwise methods performed similarly to exhaustive algorithms for subset selection, and the choice of criterion for comparing models (Akaike's information criterion, Schwarz's Bayesian information criterion or F statistics) had little effect on predictive ability. For most of the data sets, two methods based on regression trees yielded models with substantially lower predictive ability. I argue that there is no 'best' method of variable selection and that any of the regression-based approaches discussed here is capable of yielding useful predictive models.

Variable selection methods are very popular, especially in the field of big data with large predictors. These procedures improve the accuracy and performance of the model by eliminating irrelevant and redundant variables. The main contribution of this study is to couple a logit model with a novel variable selection approach, "Stability Competitive Adaptive Re-weighted Sampling" to address binary response. The efficiency of the proposed method is compared with the traditional logistic regression model based on eight model assessment criteria over real data from sexually transmitted infections in Indian men. Due to higher stability, the proposed method outperformed having a lower Akaike information criterion, and the Bayesian information criterion, as well as higher R-squared measures. The finally selected proposed model identified essential information regarding sexually transmitted infections in India for policymakers.

Evaluation of the effect of chance correlations on variable selection using Partial Least Squares-Discriminant Analysis

In this paper, we are interested in investigating the causal relationships among futures sugar prices in the Zhengzhou futures exchange market (ZF), the spot sugar prices in Zhengzhou (ZS) and the futures sugar prices in New York futures exchange market (NF). A useful tool called Bayesian network is introduced to analyze the problem. Since there are only three variables in our Bayesian network, the algorithm is straightforward: we display all the 25 possible network structures and adopt certain scoring metrics to evaluate them. We applied five different scoring metrics in total. Firstly, for each metric, we obtained 24 scores, each calculated from one of the 24 possible structures i.e. a Directed Acyclic Graph (DAG). Then we eliminated the network structure which represents the independence of the three variables according to our prior knowledge concerning the futures sugar market. After that, the optimal network structure which implies the causal relationships was selected according to the corresponding scoring metric. Finally, after comparing the results from different scoring metrics, we obtained the relatively affirmative conclusion that ZS causes ZF from both the Bayesian Dirichlet (BD) metric, Bayesian Dirichlet-Akaike Information Criterion (BD-AIC) metric, Bayesian Dirichlet-Bayesian Information Criterion (BD-BIC) metric and Bayesian Information Criterion (BIC) metric. The conclusions that NF causes ZF and ZF causes ZS from the Akaike Information Criterion (AIC) metric and ZF causes ZS from the BIC metric were useful and significant to our investigation.

We present a new Stata program, vselect, that helps users perform variable selection after performing a linear regression. Options for stepwise methods such as forward selection and backward elimination are provided. The user may specify Mallows's Cp, Akaike's information criterion, Akaike's corrected information criterion, Bayesian information criterion, or R2 adjusted as the information criterion for the selection. When the user specifies the best subset option, the leaps-and-bounds algorithm (Furnival and Wilson, Technometrics 16: 499–511) is used to determine the best subsets of each predictor size. All the previously mentioned information criteria are reported for each of these subsets. We also provide options for doing variable selection only on certain predictors (as in [R] nestreg) and support for weighted linear regression. All options are demonstrated on real datasets with varying numbers of predictors.

Most model selection procedures use a fixed penalty penalizing an increase in the size of a model. These nonadaptive selection procedures perform well only in one type of situation. For instance, Bayesian information criterion (BIC) with a large penalty performs well for “small” models and poorly for “large” models, and Akaike's information criterion (AIC) does just the opposite. This article proposes an adaptive model selection procedure that uses a data-adaptive complexity penalty based on a concept of generalized degrees of freedom. The proposed procedure, combining the benefit of a class of nonadaptive procedures, approximates the best performance of this class of procedures across a variety of different situations. This class includes many well-known procedures, such as AIC, BIC, Mallows's Cp, and risk inflation criterion (RIC). The proposed procedure is applied to wavelet thresholding in nonparametric regression and variable selection in least squares regression. Simulation results and an asymptotic analysis support the effectiveness of the proposed procedure.

Variable selection is fundamental to high dimensional statistical modeling, and many approaches have been proposed. However, existing variable selection methods do not perform well in presence of outliers in response variable or/and covariates. In order to ensure a high probability of correct selection and efficient parameter estimation, we investigate a robust variable selection method based on a modified Huber's function with an exponential squared loss tail. We also prove that the proposed method has oracle properties. Furthermore, we carry out simulation studies to evaluate the performance of the proposed method for both p<n and p>n. Our simulation results indicate that the proposed method is efficient and robust against outliers and heavy-tailed distributions. Finally, a real dataset from an air pollution mortality study is used to illustrate the proposed method.

Time series study is of paramount importance especially in developing countries such as India. The objective of this study is to forecast the number of dengue cases by applying time series models. Furthermore, this study also aims to uncover the correlation of ecological variables with dengue cases using a p value test. The ecological variables considered are the rainfall (with a 2-month lag period) and average relative humidity. An evaluation of host factors such as age and gender is done with the number of cases to determine the most affected strata. Monthly data on dengue cases from 2014 to 2017 are taken from the integrated disease surveillance programme, Government of India, and data on ecological variables, humidity and rainfall are taken from www.Indiastat.com . A 2-month lag period is considered for rainfall. These data were analysed by the autoregressive integrated moving average (ARIMA) and exponential smoothing model (ES). Predictive performance of these techniques was assessed by evaluation parameters such as the Akaike information criterion (AIC), Bayesian information criterion (BIC), mean absolute error and mean absolute scaled error. Applying time series models was useful in the prediction. There were many more males than females according to the data for 2014–2017. The Akaike information criterion (AIC), Bayesian information criterion (BIC) and mean absolute scaled error for ARIMA are 48.6526, 448.333333, 449.30008, 453.08 and 0.507826, respectively. Based on model indices such as the Akaike information criterion (AIC), Bayesian information criterion (BIC), mean absolute error and mean absolute scaled error, the ARIMA performed better than the ES model. Rainfall at the 2-month lag was correlated with the number of cases.

Modern regularization and variable selection methods, such as least absolute shrinkage and selection operator (lasso) and Bayesian variable selection, are important tools for psychological researchers to reduce the risk of overfitting, improve prediction in future samples, and increase model interpretability. Although missing data are common in psychological data, it is not straightforward to combine principled methods for addressing missing data with these modern variable selection methods. This challenge is well illustrated in a recent article by Gunn et al. (2023) with a comparison of three approaches for combining lasso with multiple imputation to address missing data. Each of the surveyed approaches results in markedly different results in terms of predictors selected. Their findings underscore limitations of the lasso for the purpose of variable selection. In this article, we show how to implement a Bayesian variable selection method, stochastic search variable selection (SSVS), with multiply imputed data. SSVS is a principled and consistent method for variable selection, and we demonstrate advantages relative to lasso in an example data set and simulation study. It is straightforward to apply an ITS strategy for SSVS using existing software. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

SummaryThis paper considers a linear regression model with regression parameter vector β. The parameter of interest is θ= aTβ where a is specified. When, as a first step, a data‐based variable selection (e.g. minimum Akaike information criterion) is used to select a model, it is common statistical practice to then carry out inference about θ, using the same data, based on the (false) assumption that the selected model had been provided a priori. The paper considers a confidence interval for θ with nominal coverage 1 ‐ α constructed on this (false) assumption, and calls this the naive 1 ‐ α confidence interval. The minimum coverage probability of this confidence interval can be calculated for simple variable selection procedures involving only a single variable. However, the kinds of variable selection procedures used in practice are typically much more complicated. For the real‐life data presented in this paper, there are 20 variables each of which is to be either included or not, leading to 220 different models. The coverage probability at any given value of the parameters provides an upper bound on the minimum coverage probability of the naive confidence interval. This paper derives a new Monte Carlo simulation estimator of the coverage probability, which uses conditioning for variance reduction. For these real‐life data, the gain in efficiency of this Monte Carlo simulation due to conditioning ranged from 2 to 6. The paper also presents a simple one‐dimensional search strategy for parameter values at which the coverage probability is relatively small. For these real‐life data, this search leads to parameter values for which the coverage probability of the naive 0.95 confidence interval is 0.79 for variable selection using the Akaike information criterion and 0.70 for variable selection using Bayes information criterion, showing that these confidence intervals are completely inadequate.

Variable selection can improve the robustness and prediction accuracy of partial least squares (PLS) regression models and decrease the calculation time by selecting the optimal subset of variables in multivariate calibration. In this study, the performance of two variable selection methods for wavelength interval and individual wavelength coupled with partial least squares regression are investigated by employing the experimental data of asiaticoside (AS) and madecassoside (MS) contents in centella total glucosides (CTG) and a public dataset of corn. The studied variable selection methods include interval partial least squares regression ( iPLS), backward interval partial least squares ( biPLS), synergy interval partial least squares regression ( siPLS), competitive adaptive reweighted sampling (CARS), uninformative variable elimination (UVE) and variable importance in projection (VIP). The results show that the implementation of variable selection methods improved the performance of the model compared with full-spectrum modeling. All variable selection methods improved the prediction of AS or MS contents in CTG. When latent variables for PLS models are less than 10 in the practical application, the RPD value of AS models by iPLS method is 7.5, and the RPD value of MS models by biPLS method is 2.9. The results of wavelength interval selection are better than individual wavelength selection, especially for iPLS and biPLS. The same results were obtained with the public data for moisture in corn, and the RPD value of biPLS model of moisture is 1.6. Therefore, the wavelength interval selection methods, such as iPLS or biPLS, are appropriate for improving the PLS model’s accuracy and robustness to determine the target components’ contents in solid samples.

Variable (wavelength) selection is critical in the multivariate calibration of spectra that improves prediction performance and provides clearer interpretation. Most variable selection methods use Y information (chemical reference of interest) to evaluating the importance of variables. In this work, a novel variable selection is proposed that requires no reference information. Replicate spectra of a sample under various acquisition conditions were used with an index to evaluate the importance of variables. The results show that the proposed method achieves better results than the original partial least squares model with a few available calibration samples. When a large number of calibration samples are available, this method provided comparable results to the state-of-the-art variable selection methods while reduced the risk of model over-fitting.

This paper introduces a SAS/IML program to select among the multivariate model candidates based on a few well-known multivariate model selection criteria. Stepwise regression and all-possible-regression are considered. The program is user friendly and requires the user to paste or read the data at the beginning of the module, include the names of the dependent and independent variables (the y's and the x's), and then run the module. The program produces the multivariate candidate models based on the following criteria: Forward Selection, Forward Stepwise Regression, Backward Elimination, Mean Square Error, Coefficient of Multiple Determination, Adjusted Coefficient of Multiple Determination, Akaike's Information Criterion, the Corrected Form of Akaike's Information Criterion, Hannan and Quinn Information Criterion, the Corrected Form of Hannan and Quinn (HQc) Information Criterion, Schwarz's Criterion, and Mallow's PC. The output also constitutes detailed as well as summarized results.

BackgroundThe redundancy of information is becoming a critical issue for epidemiologists. High-dimensional datasets require new effective variable selection methods to be developed. This study implements an advanced evolutionary variable selection method which is applied for cardiovascular predictive modeling. The epidemiological follow-up study KIHD (Kuopio Ischemic Heart Disease Risk Factor Study) was used to compare the designed variable selection method based on an evolutionary search with conventional stepwise selection. The sample contains in total 433 predictor variables and a response variable indicating incidents of cardiovascular diseases for 1465 study subjects.ResultsThe effectiveness of variable selection methods was investigated in combination with two models: Generalized Linear Logistic Regression and Support Vector Machine. We managed to decrease the number of variables from 433 to 38 and save the predictive ability of the models used. Their performance was evaluated with an F-score metric. At most, we gained 65.6% and 67.4% of the F-score before and after variable selection respectively. All the results were averaged over 5-folds of a cross-validation procedure.ConclusionsThe presented evolutionary variable selection method allows a reduced set of variables to be chosen which are relevant to predicting cardiovascular diseases. A reference list of the most meaningful variables is introduced to be used as a basis for new epidemiological studies. In general, the multicollinearity of variables enables different combinations of predictors to be used and the same performance of models to be attained.

This study aimed to explore the latent factors associated to functional dentition (FD), quantify their clustering across probability levels, and derive precision prevention strategies. A cross-sectional study of 423 adults aged 65-74 were conducted using data from a 2021 national surveillance project in China. After preliminary selection of potential variables through multivariable logistic regression, latent class analysis (LCA) was performed in R. Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and entropy determined the optimal number of latent classes. Bootstrap resampling and a confusion matrix assessed model performance. Participants were classified into a low-probability group (Class 1, N = 136, 32.2%) and a high-probability group (Class 2, N = 287, 67.8%), representing the best fit (AIC = 2700.36, BIC = 2744.88, Entropy = 0.733). Compared to Class 2, Class 1 exhibited lower toothbrushing frequency and oral health knowledge level, but higher proportions of root caries, AL > 5mm, and unrestored tooth loss. The model showed good stability (average agreement rate = 0.867) and strong predictive performance (Kappa = 0.806, Accuracy = 0.915, Sensitivity = 0.862, Specificity = 0.940, Precision = 0.875, F1-score = 0.869). LCA identified distinct subgroups at varying FD probability and quantified clustered associated factors, supporting the accurate identification of high-risk populations and the development of targeted prevention strategies.