# Noah Simon

#### 11 packages on CRAN

This package searches for marginal interactions in a binary response model. Interact uses permutation methods to estimate false discovery rates for these marginal interactions and has some, limited visualization capabilities

Fit a regularized generalized linear model via penalized maximum likelihood. The model is fit for a path of values of the penalty parameter. Fits linear, logistic and Cox models.

Fits Cox model via stochastic gradient descent. This implementation avoids computational instability of the standard Cox Model when dealing large datasets. Furthermore, it scales up with large datasets that do not fit the memory. It also handles large sparse datasets using proximal stochastic gradient descent algorithm. For more details about the method, please see Aliasghar Tarkhan and Noah Simon (2020) <arXiv:2003.00116v2>.

Develop and evaluate treatment rules based on: (1) the standard indirect approach of split-regression, which fits regressions separately in both treatment groups and assigns an individual to the treatment option under which predicted outcome is more desirable; (2) the direct approach of outcome-weighted-learning proposed by Yingqi Zhao, Donglin Zeng, A. John Rush, and Michael Kosorok (2012) <doi:10.1080/01621459.2012.695674>; (3) the direct approach, which we refer to as direct-interactions, proposed by Shuai Chen, Lu Tian, Tianxi Cai, and Menggang Yu (2017) <doi:10.1111/biom.12676>. Please see the vignette for a walk-through of how to start with an observational dataset whose design is understood scientifically and end up with a treatment rule that is trustworthy statistically, along with an estimation of rule benefit in an independent sample.

Extremely efficient procedures for fitting the entire lasso or elastic-net regularization path for linear regression, logistic and multinomial regression models, Poisson regression, Cox model, multiple-response Gaussian, and the grouped multinomial regression. There are two new and important additions. The family argument can be a GLM family object, which opens the door to any programmed family. This comes with a modest computational cost, so when the built-in families suffice, they should be used instead. The other novelty is the relax option, which refits each of the active sets in the path unpenalized. The algorithm uses cyclical coordinate descent in a path-wise fashion, as described in the papers listed in the URL below.

Implementation of the convex hierarchical testing (CHT) procedure introduced in Bien, Simon, and Tibshirani (2015) Convex Hierarchical Testing of Interactions. Annals of Applied Statistics. Vol. 9, No. 1, 27-42.

Efficient implementations for Sorted L-One Penalized Estimation (SLOPE): generalized linear models regularized with the sorted L1-norm (Bogdan et al. (2015) <doi:10/gfgwzt>) or, equivalently, ordered weighted L1-norm (OWL). Supported models include ordinary least-squares regression, binomial regression, multinomial regression, and Poisson regression. Both dense and sparse predictor matrices are supported. In addition, the package features predictor screening rules that enable fast and efficient solutions to high-dimensional problems.

Proximal gradient descent solver for the operators lasso, group lasso, and sparse-group lasso. The implementation involves backtracking line search and warm starts. Input data needs to be clustered/grouped for the (sparse-)group lasso before calling these algorithms.

Efficient implementations for Sorted L-One Penalized Estimation (SLOPE): generalized linear models regularized with the sorted L1-norm (Bogdan et al. (2015) <doi:10/gfgwzt>). Supported models include ordinary least-squares regression, binomial regression, multinomial regression, and Poisson regression. Both dense and sparse predictor matrices are supported. In addition, the package features predictor screening rules that enable fast and efficient solutions to high-dimensional problems.

Fit a trio model via penalized maximum likelihood. The model is fit for a path of values of the penalty parameter. This package is based on Noah Simon, et al. (2011) <doi:10.1080/10618600.2012.681250>.

Calculate point estimates of and valid confidence intervals for nonparametric, algorithm-agnostic variable importance measures in high and low dimensions, using flexible estimators of the underlying regression functions. For more information about the methods, please see Williamson et al. (Biometrics, 2020), Williamson et al. (arXiv, 2020+) <arXiv:2004.03683>, and Williamson and Feng (ICML, 2020) <arXiv:>.