Variable selection for BART: An application to gene regulation

Bleich, Justin; Kapelner, Adam; George, Edward I.; Jensen, Shane T.

doi:10.1214/14-aoas755

Cited by 120 publications

(166 citation statements)

References 39 publications

Supporting

Mentioning

164

Contrasting

Order By: Relevance

“…To establish estimates of what variables matter and to gain a (rough) sense of the structure of each BART model, we calculate each variable's inclusion proportion. This quantity measures the number of times a specific variable was used in BART's tree models, divided by the total number of variables used in all of the BART's tree models (Chipman et al, 2010; Kaplener & Bleich, 2014) The more often a variable is used in predicting the response, the higher its inclusion proportion will be. Plots of the top 20 variables by inclusion proportion (for each model) are included in the supplemental materials. Note that, for the Sample I and Sample II self-rated health and depression models, social-psychological features (e.g.…”

mentioning

confidence: 99%

The consequences of contact with the criminal justice system for health in the transition to adulthood

Esposito

Lee

Hicken

et al. 2017

LLCS

View full text Add to dashboard Cite

A rapidly growing literature has documented the adverse social, economic and, recently, health impacts of experiencing incarceration in the United States. Despite the insights that this work has provided in consistently documenting the deleterious effects of incarceration, little is known about the specific timing of criminal justice contact and early health consequences during the transition from adolescence to adulthood—a critical period in the life course, particularly for the development of poor health. Previous literature on the role of incarceration has also been hampered by the difficulties of parsing out the influence that incarceration exerts on health from the social and economic confounding forces that are linked to both criminal justice contact and health. This paper addresses these two gaps in the literature by examining the association between incarceration and health in the United States during the transition to adulthood, and by using an analytic approach that better isolates the association of incarceration with health from the multitude of confounders which could be alternatively driving this association. In this endeavor, we make use of variable-rich data from The National Longitudinal Study of Adolescent to Adult Health (n = 10,785) and a non-parametric Bayesian machine learning technique- Bayesian Additive Regression Trees. Our results suggest that the experience of incarceration at this stage of the life course increases the probability of depression, adversely affects the perception of general health status, but has no effect on the probability of developing hypertension in early adulthood. These findings signal that incarceration in emerging adulthood is an important stressor that can have immediate implications for mental and general health in early adulthood, and may help to explain long lasting implications incarceration has for health across the life course.

show abstract

mentioning

confidence: 99%

The consequences of contact with the criminal justice system for health in the transition to adulthood

Esposito

Lee

Hicken

et al. 2017

LLCS

View full text Add to dashboard Cite

show abstract

“…Bleich et al (2014) discuss various ways of choosing appropriate thresholds for the variable inclusion proportions to use BART for variable selection such as the “local” and “global max threshold” options. Both of these involve permuting the response variable and running the BART model a number of times.…”

Section: Resultsmentioning

confidence: 99%

Bayesian Additive Regression Trees using Bayesian model averaging

et al. 2017

View full text Add to dashboard Cite

Bayesian Additive Regression Trees (BART) is a statistical sum of trees model. It can be considered a Bayesian version of machine learning tree ensemble methods where the individual trees are the base learners. However for datasets where the number of variables p is large the algorithm can become inefficient and computationally expensive. Another method which is popular for high dimensional data is random forests, a machine learning algorithm which grows trees using a greedy search for the best split points. However its default implementation does not produce probabilistic estimates or predictions. We propose an alternative fitting algorithm for BART called BART-BMA, which uses Bayesian Model Averaging and a greedy search algorithm to obtain a posterior distribution more efficiently than BART for datasets with large p. BART-BMA incorporates elements of both BART and random forests to offer a model-based algorithm which can deal with high-dimensional data. We have found that BART-BMA can be run in a reasonable time on a standard laptop for the “small n large p” scenario which is common in many areas of bioinformatics. We showcase this method using simulated data and data from two real proteomic experiments, one to distinguish between patients with cardiovascular disease and controls and another to classify aggressive from non-aggressive prostate cancer. We compare our results to their main competitors. Open source code written in R and Rcpp to run BART-BMA can be found at: https://github.com/BelindaHernandez/BART-BMA.git

show abstract

“…Additional variable selection methods using the average use per splitting rule are also available in and described in Bleich et al . . These include using permutation sampling to determine an appropriate threshold for the average use per splitting rule based on a null distribution, which would help identify which variables are truly important.…”

Section: Regressionmentioning

confidence: 99%

Nonparametric survival analysis using Bayesian Additive Regression Trees (BART)

Sparapani

Logan

McCulloch

et al. 2016

Statistics in Medicine

138

125

View full text Add to dashboard Cite

Bayesian additive regression trees (BART) provide a framework for flexible nonparametric modeling of relationships of covariates to outcomes. Recently, BART models have been shown to provide excellent predictive performance, for both continuous and binary outcomes, and exceeding that of its competitors. Software is also readily available for such outcomes. In this article we introduce modeling that extends the usefulness of BART in medical applications by addressing needs arising in survival analysis. Simulation studies of one-sample and two-sample scenarios, in comparison with long-standing traditional methods, establish face validity of the new approach. We then demonstrate the model’s ability to accommodate data from complex regression models with a simulation study of a nonproportional hazards scenario with crossing survival functions, and survival function estimation in a scenario where hazards are multiplicatively modified by a highly nonlinear function of the covariates. Using data from a recently published study of patients undergoing hematopoietic stem cell transplantation, we illustrate the use and some advantages of the proposed method in medical investigations.

show abstract

Variable selection for BART: An application to gene regulation

Cited by 120 publications

References 39 publications

The consequences of contact with the criminal justice system for health in the transition to adulthood

The consequences of contact with the criminal justice system for health in the transition to adulthood

Bayesian Additive Regression Trees using Bayesian model averaging

Nonparametric survival analysis using Bayesian Additive Regression Trees (BART)

Contact Info

Product

Resources

About