Search

Rigorous external validation is crucial for assessing the generalizability of prediction models, particularly by evaluating their discrimination (AUROC) on new data. This often involves comparing a new model's AUROC to that of an established reference model. However, many studies rely on arbitrary rules of thumb for sample size calculations, often resulting in underpowered analyses and unreliable conclusions. This paper reviews crucial concepts for accurate sample size determination in AUROC-based external validation studies, making the theory and practice more accessible to researchers and clinicians. We introduce powerROC, an open-source web tool designed to simplify these calculations, enabling both the evaluation of a single model and the comparison of two models. The tool offers guidance on selecting target precision levels and employs flexible approaches, leveraging either pilot data or user-defined probability distributions. We illustrate powerROC's utility through a case study on hospital mortality prediction using the MIMIC database.

This paper proposes a sparse regression method that continuously interpolates between Forward Stepwise selection (FS) and the LASSO. When tuned appropriately, our solutions are much sparser than typical LASSO fits but, unlike FS fits, benefit from the stabilizing effect of shrinkage. Our method, Adaptive Forward Stepwise Regression (AFS) addresses this need for sparser models with shrinkage. We show its connection with boosting via a soft-thresholding viewpoint and demonstrate the ease of adapting the method to classification tasks. In both simulations and real data, our method has lower mean squared error and fewer selected features across multiple settings compared to popular sparse modeling procedures.

Many risk-sensitive applications require well-calibrated prediction sets over multiple, potentially correlated target variables, for which the prediction algorithm may report correlated non-conformity scores. In this work, we treat the scores as random vectors and aim to construct the prediction set accounting for their joint correlation structure. Drawing from the rich literature on multivariate quantiles and semiparametric statistics, we propose an algorithm to estimate the $1-\alpha$ quantile of the scores, where $\alpha$ is the user-specified miscoverage rate. In particular, we flexibly estimate the joint cumulative distribution function (CDF) of the scores using nonparametric vine copulas and improve the asymptotic efficiency of the quantile estimate using its influence function. The vine decomposition allows our method to scale well to a large number of targets. We report desired coverage and competitive efficiency on a range of real-world regression problems, including those with missing-at-random labels in the calibration set., Comment: 9 pages (+12 appendix), 11 figures, submitted to AISTATS 2025

We propose Cooperative Component Analysis (CoCA), a new method for unsupervised multi-view analysis: it identifies the component that simultaneously captures significant within-view variance and exhibits strong cross-view correlation. The challenge of integrating multi-view data is particularly important in biology and medicine, where various types of "-omic" data, ranging from genomics to proteomics, are measured on the same set of samples. The goal is to uncover important, shared signals that represent underlying biological mechanisms. CoCA combines an approximation error loss to preserve information within data views and an "agreement penalty" to encourage alignment across data views. By balancing the trade-off between these two key components in the objective, CoCA has the property of interpolating between the commonly-used principal component analysis (PCA) and canonical correlation analysis (CCA) as special cases at the two ends of the solution path. CoCA chooses the degree of agreement in a data-adaptive manner, using a validation set or cross-validation to estimate test error. Furthermore, we propose a sparse variant of CoCA that incorporates the Lasso penalty to yield feature sparsity, facilitating the identification of key features driving the observed patterns. We demonstrate the effectiveness of CoCA on simulated data and two real multiomics studies of COVID-19 and ductal carcinoma in situ of breast. In both real data applications, CoCA successfully integrates multiomics data, extracting components that are not only consistently present across different data views but also more informative and predictive of disease progression. CoCA offers a powerful framework for discovering important shared signals in multi-view data, with the potential to uncover novel insights in an increasingly multi-view data world.

Single-cell datasets often lack individual cell labels, making it challenging to identify cells associated with disease. To address this, we introduce Mixture Modeling for Multiple Instance Learning (MMIL), an expectation maximization method that enables the training and calibration of cell-level classifiers using patient-level labels. Our approach can be used to train e.g. lasso logistic regression models, gradient boosted trees, and neural networks. When applied to clinically-annotated, primary patient samples in Acute Myeloid Leukemia (AML) and Acute Lymphoblastic Leukemia (ALL), our method accurately identifies cancer cells, generalizes across tissues and treatment timepoints, and selects biologically relevant features. In addition, MMIL is capable of incorporating cell labels into model training when they are known, providing a powerful framework for leveraging both labeled and unlabeled data simultaneously. Mixture Modeling for MIL offers a novel approach for cell classification, with significant potential to advance disease understanding and management, especially in scenarios with unknown gold-standard labels and high dimensionality., Comment: Erin Craig and Timothy Keyes contributed equally to this work

Recent genome-wide association studies (GWAS) have uncovered the genetic basis of complex traits, but show an under-representation of non-European descent individuals, underscoring a critical gap in genetic research. Here, we assess whether we can improve disease prediction across diverse ancestries using multiomic data. We evaluate the performance of Group-LASSO INTERaction-NET (glinternet) and pretrained lasso in disease prediction focusing on diverse ancestries in the UK Biobank. Models were trained on data from White British and other ancestries and validated across a cohort of over 96,000 individuals for 8 diseases. Out of 96 models trained, we report 16 with statistically significant incremental predictive performance in terms of ROC-AUC scores (p-value < 0.05), found for diabetes, arthritis, gall stones, cystitis, asthma and osteoarthritis. For the interaction and pretrained models that outperformed the baseline, the PRS score was the primary driver behind prediction. Our findings indicate that both interaction terms and pre-training can enhance prediction accuracy but for a limited set of diseases and moderate improvements in accuracy

Pretraining is a popular and powerful paradigm in machine learning to pass information from one model to another. As an example, suppose one has a modest-sized dataset of images of cats and dogs, and plans to fit a deep neural network to classify them from the pixel features. With pretraining, we start with a neural network trained on a large corpus of images, consisting of not just cats and dogs but hundreds of other image types. Then we fix all of the network weights except for the top layer(s) (which makes the final classification) and train (or "fine tune") those weights on our dataset. This often results in dramatically better performance than the network trained solely on our smaller dataset. In this paper, we ask the question "Can pretraining help the lasso?". We develop a framework for the lasso in which a model is fit to a large dataset, and then fine-tuned using a smaller dataset. This latter dataset can be a subset of the original dataset, or it can be a dataset with a different but related outcome. This framework has a wide variety of applications, including stratified models, multinomial responses, multi-response models, conditional average treatment estimation and even gradient boosting. In the stratified model setting, the pretrained lasso pipeline estimates the coefficients common to all groups at the first stage, and then estimates the group-specific coefficients at the second "fine-tuning" stage. We show that under appropriate assumptions, the support recovery rate of the common coefficients is superior to that of the usual lasso trained only on individual groups. This separate identification of common and individual coefficients can also be useful for scientific understanding.

Inference for prediction errors is critical in time series forecasting pipelines. However, providing statistically meaningful uncertainty intervals for prediction errors remains relatively under-explored. Practitioners often resort to forward cross-validation (FCV) for obtaining point estimators and constructing confidence intervals based on the Central Limit Theorem (CLT). The naive version assumes independence, a condition that is usually invalid due to time correlation. These approaches lack statistical interpretations and theoretical justifications even under stationarity. This paper systematically investigates uncertainty intervals for prediction errors in time series forecasting. We first distinguish two key inferential targets: the stochastic test error over near future data points, and the expected test error as the expectation of the former. The stochastic test error is often more relevant in applications needing to quantify uncertainty over individual time series instances. To construct prediction intervals for the stochastic test error, we propose the quantile-based forward cross-validation (QFCV) method. Under an ergodicity assumption, QFCV intervals have asymptotically valid coverage and are shorter than marginal empirical quantiles. In addition, we also illustrate why naive CLT-based FCV intervals fail to provide valid uncertainty intervals, even with certain corrections. For non-stationary time series, we further provide rolling intervals by combining QFCV with adaptive conformal prediction to give time-average coverage guarantees. Overall, we advocate the use of QFCV procedures and demonstrate their coverage and efficiency through simulations and real data examples., Comment: 35 pages, 17 figures

Multiomics data fusion integrates diverse data modalities, ranging from transcriptomics to proteomics, to gain a comprehensive understanding of biological systems and enhance predictions on outcomes of interest related to disease phenotypes and treatment responses. Cooperative learning, a recently proposed method, unifies the commonly-used fusion approaches, including early and late fusion, and offers a systematic framework for leveraging the shared underlying relationships across omics to strengthen signals. However, the challenge of acquiring large-scale labeled data remains, and there are cases where multiomics data are available but in the absence of annotated labels. To harness the potential of unlabeled multiomcis data, we introduce semi-supervised cooperative learning. By utilizing an "agreement penalty", our method incorporates the additional unlabeled data in the learning process and achieves consistently superior predictive performance on simulated data and a real multiomics study of aging. It offers an effective solution to multiomics data fusion in settings with both labeled and unlabeled data and maximizes the utility of available data resources, with the potential of significantly improving predictive models for diagnostics and therapeutics in an increasingly multiomics world., Comment: The 2023 ICML Workshop on Machine Learning for Multimodal Healthcare Data. arXiv admin note: text overlap with arXiv:2112.12337

Hierarchical clustering is a popular method for identifying distinct groups in a dataset. The most commonly used method for pruning a dendrogram is via a single horizontal cut. In this paper, we propose a new technique "weakest link optimal pruning". We prove its superiority over horizontal pruning and provide some examples illustrating how the two methods can behave quite differently.

The Cox proportional hazards model is a canonical method in survival analysis for prediction of the life expectancy of a patient given clinical or genetic covariates -- it is a linear model in its original form. In recent years, several methods have been proposed to generalize the Cox model to neural networks, but none of these are both numerically correct and computationally efficient. We propose FastCPH, a new method that runs in linear time and supports both the standard Breslow and Efron methods for tied events. We also demonstrate the performance of FastCPH combined with LassoNet, a neural network that provides interpretability through feature sparsity, on survival datasets. The final procedure is efficient, selects useful covariates and outperforms existing CoxPH approaches.

In supervised learning, the estimation of prediction error on unlabeled test data is an important task. Existing methods are usually built on the assumption that the training and test data are sampled from the same distribution, which is often violated in practice. As a result, traditional estimators like cross-validation (CV) will be biased and this may result in poor model selection. In this paper, we assume that we have a test dataset in which the feature values are available but not the outcome labels, and focus on a particular form of distributional shift called "covariate shift". We propose an alternative method based on parametric bootstrap of the target of conditional error. Empirically, our method outperforms CV for both simulation and real data example across different modeling tasks.

Forecasting methodologies have always attracted a lot of attention and have become an especially hot topic since the beginning of the COVID-19 pandemic. In this paper we consider the problem of multi-period forecasting that aims to predict several horizons at once. We propose a novel approach that forces the prediction to be "smooth" across horizons and apply it to two tasks: point estimation via regression and interval prediction via quantile regression. This methodology was developed for real-time distributed COVID-19 forecasting. We illustrate the proposed technique with the CovidCast dataset as well as a small simulation example.

Standard regression adjustment gives inconsistent estimates of causal effects when there are time-varying treatment effects and time-varying covariates. Loosely speaking, the issue is that some covariates are post-treatment variables because they may be affected by prior treatment status, and regressing out post-treatment variables causes bias. More precisely, the bias is due to certain non-confounding latent variables that create colliders in the causal graph. These latent variables, which we call phantoms, do not harm the identifiability of the causal effect, but they render naive regression estimates inconsistent. Motivated by this, we ask: how can we modify regression methods so that they hold up even in the presence of phantoms? We develop an estimator for this setting based on regression modeling (linear, log-linear, probit and Cox regression), proving that it is consistent for a reasonable causal estimand. In particular, the estimator is a regression model fit with a simple adjustment for collinearity, making it easy to understand and implement with standard regression software. The proposed estimators are instances of the parametric g-formula, extending the regression-with-residuals approach to several canonical nonlinear models.

Out-of-bag error is commonly used as an estimate of generalisation error in ensemble-based learning models such as random forests. We present confidence intervals for this quantity using the delta-method-after-bootstrap and the jackknife-after-bootstrap techniques. These methods do not require growing any additional trees. We show that these new confidence intervals have improved coverage properties over the naive confidence interval, in real and simulated examples., Comment: 25 pages, 8 tables, 8 figures

Cross-validation (CV) is one of the most widely used techniques in statistical learning for estimating the test error of a model, but its behavior is not yet fully understood. It has been shown that standard confidence intervals for test error using estimates from CV may have coverage below nominal levels. This phenomenon occurs because each sample is used in both the training and testing procedures during CV and as a result, the CV estimates of the errors become correlated. Without accounting for this correlation, the estimate of the variance is smaller than it should be. One way to mitigate this issue is by estimating the mean squared error of the prediction error instead using nested CV. This approach has been shown to achieve superior coverage compared to intervals derived from standard CV. In this work, we generalize the nested CV idea to the Cox proportional hazards model and explore various choices of test error for this setting.

We propose a new method for supervised learning with multiple sets of features ("views"). The multiview problem is especially important in biology and medicine, where "-omics" data such as genomics, proteomics and radiomics are measured on a common set of samples. Cooperative learning combines the usual squared error loss of predictions with an "agreement" penalty to encourage the predictions from different data views to agree. By varying the weight of the agreement penalty, we get a continuum of solutions that include the well-known early and late fusion approaches. Cooperative learning chooses the degree of agreement (or fusion) in an adaptive manner, using a validation set or cross-validation to estimate test set prediction error. One version of our fitting procedure is modular, where one can choose different fitting mechanisms (e.g. lasso, random forests, boosting, neural networks) appropriate for different data views. In the setting of cooperative regularized linear regression, the method combines the lasso penalty with the agreement penalty, yielding feature sparsity. The method can be especially powerful when the different data views share some underlying relationship in their signals that can be exploited to boost the signals. We show that cooperative learning achieves higher predictive accuracy on simulated data and a real multiomics example of labor onset prediction. Leveraging aligned signals and allowing flexible fitting mechanisms for different modalities, cooperative learning offers a powerful approach to multiomics data fusion.

While there are many well-developed data science methods for classification and regression, there are relatively few methods for working with right-censored data. Here, we present "survival stacking": a method for casting survival analysis problems as classification problems, thereby allowing the use of general classification methods and software in a survival setting. Inspired by the Cox partial likelihood, survival stacking collects features and outcomes of survival data in a large data frame with a binary outcome. We show that survival stacking with logistic regression is approximately equivalent to the Cox proportional hazards model. We further recommend methods for evaluating model performance in the survival stacked setting, and we illustrate survival stacking on real and simulated data. By reframing survival problems as classification problems, we make it possible for data scientists to use well-known learning algorithms (including random forests, gradient boosting machines and neural networks) in a survival setting, and lower the barrier for flexible survival modeling.

Cross-validation is a widely-used technique to estimate prediction error, but its behavior is complex and not fully understood. Ideally, one would like to think that cross-validation estimates the prediction error for the model at hand, fit to the training data. We prove that this is not the case for the linear model fit by ordinary least squares; rather it estimates the average prediction error of models fit on other unseen training sets drawn from the same population. We further show that this phenomenon occurs for most popular estimates of prediction error, including data splitting, bootstrapping, and Mallow's Cp. Next, the standard confidence intervals for prediction error derived from cross-validation may have coverage far below the desired level. Because each data point is used for both training and testing, there are correlations among the measured accuracies for each fold, and so the usual estimate of variance is too small. We introduce a nested cross-validation scheme to estimate this variance more accurately, and we show empirically that this modification leads to intervals with approximately correct coverage in many examples where traditional cross-validation intervals fail.

Identifying heterogeneous treatment effects (HTEs) in randomized controlled trials is an important step toward understanding and acting on trial results. However, HTEs are often small and difficult to identify, and HTE modeling methods which are very general can suffer from low power. We present a method that exploits any existing relationship between illness severity and treatment effect, and identifies the "sweet spot", the contiguous range of illness severity where the estimated treatment benefit is maximized. We further compute a bias-corrected estimate of the conditional average treatment effect (CATE) in the sweet spot, and a $p$-value. Because we identify a single sweet spot and $p$-value, we believe our method to be straightforward to interpret and actionable: results from our method can inform future clinical trials and help clinicians make personalized treatment recommendations., Comment: 14 pages, 14 figures

In some supervised learning settings, the practitioner might have additional information on the features used for prediction. We propose a new method which leverages this additional information for better prediction. The method, which we call the feature-weighted elastic net ("fwelnet"), uses these "features of features" to adapt the relative penalties on the feature coefficients in the elastic net penalty. In our simulations, fwelnet outperforms the lasso in terms of test mean squared error and usually gives an improvement in true positive rate or false positive rate for feature selection. We also apply this method to early prediction of preeclampsia, where fwelnet outperforms the lasso in terms of 10-fold cross-validated area under the curve (0.86 vs. 0.80). We also provide a connection between fwelnet and the group lasso and suggest how fwelnet might be used for multi-task learning.

We study the assessment of the accuracy of heterogeneous treatment effect (HTE) estimation, where the HTE is not directly observable so standard computation of prediction errors is not applicable. To tackle the difficulty, we propose an assessment approach by constructing pseudo-observations of the HTE based on matching. Our contributions are three-fold: first, we introduce a novel matching distance derived from proximity scores in random forests; second, we formulate the matching problem as an average minimum-cost flow problem and provide an efficient algorithm; third, we propose a match-then-split principle for the assessment with cross-validation. We demonstrate the efficacy of the assessment approach on synthetic data and data generated from a real dataset.

In their study, McKinney et al. showed the high potential of artificial intelligence for breast cancer screening. However, the lack of detailed methods and computer code undermines its scientific value. We identify obstacles hindering transparent and reproducible AI research as faced by McKinney et al and provide solutions with implications for the broader field.

Sparse generalized additive models (GAMs) are an extension of sparse generalized linear models which allow a model's prediction to vary non-linearly with an input variable. This enables the data analyst build more accurate models, especially when the linearity assumption is known to be a poor approximation of reality. Motivated by reluctant interaction modeling (Yu et al. 2019), we propose a multi-stage algorithm, called $\textit{reluctant generalized additive modeling (RGAM)}$, that can fit sparse generalized additive models at scale. It is guided by the principle that, if all else is equal, one should prefer a linear feature over a non-linear feature. Unlike existing methods for sparse GAMs, RGAM can be extended easily to binary, count and survival data. We demonstrate the method's effectiveness on real and simulated examples., Comment: Change of method name, R package now available

We develop a baseline-adjusted C-index to evaluate fitted Cox proportional hazard models. This metric is particularly useful in evaluating stratified Cox models, as well as model selection using cross validation. Our metric is able to compare all pairs of comparable individuals in strata or cross validation folds, as opposed to only pairs within the same stratum/folds. We demonstrate through simulations and real data applications that the baseline adjusted C-index is more stable and that it selects better model in high-dimensional $L^1$ penalized Cox regression.

In this paper, we explore a method for treating survival analysis as a classification problem. The method uses a "stacking" idea that collects the features and outcomes of the survival data in a large data frame, and then treats it as a classification problem. In this framework, various statistical learning algorithms (including logistic regression, random forests, gradient boosting machines and neural networks) can be applied to estimate the parameters and make predictions. For stacking with logistic regression, we show that this approach is approximately equivalent to the Cox proportional hazards model with both theoretical analysis and simulation studies. For stacking with other machine learning algorithms, we show through simulation studies that our method can outperform Cox proportional hazards model in terms of estimated survival curves. This idea is not new, but we believe that it should be better known by statistiicians and other data scientists., Comment: 15 pages, 8 figures, 3 Tables

Much work has been done recently to make neural networks more interpretable, and one obvious approach is to arrange for the network to use only a subset of the available features. In linear models, Lasso (or $\ell_1$-regularized) regression assigns zero weights to the most irrelevant or redundant features, and is widely used in data science. However the Lasso only applies to linear models. Here we introduce LassoNet, a neural network framework with global feature selection. Our approach enforces a hierarchy: specifically a feature can participate in a hidden unit only if its linear representative is active. Unlike other approaches to feature selection for neural nets, our method uses a modified objective function with constraints, and so integrates feature selection with the parameter learning directly. As a result, it delivers an entire regularization path of solutions with a range of feature sparsity. On systematic experiments, LassoNet significantly outperforms state-of-the-art methods for feature selection and regression. The LassoNet method uses projected proximal gradient descent, and generalizes directly to deep networks. It can be implemented by adding just a few lines of code to a standard neural network., Comment: 18 pages, 10 fg. arXiv admin note: text overlap with arXiv:1901.09346 by other authors