haldensify - Highly Adaptive Lasso Conditional Density Estimation
An algorithm for flexible conditional density estimation based on application of pooled hazard regression to an artificial repeated measures dataset constructed by discretizing the support of the outcome variable. To facilitate flexible estimation of the conditional density, the highly adaptive lasso, a non-parametric regression function shown to estimate cadlag (RCLL) functions at a suitably fast convergence rate, is used. The use of pooled hazards regression for conditional density estimation as implemented here was first described for by Díaz and van der Laan (2011) <doi:10.2202/1557-4679.1356>. Building on the conditional density estimation utilities, non-parametric inverse probability weighted (IPW) estimators of the causal effects of additive modified treatment policies are implemented, using conditional density estimation to estimate the generalized propensity score. Non-parametric IPW estimators based on this can be coupled with sieve estimation (undersmoothing) of the generalized propensity score to attain the semi-parametric efficiency bound (per Hejazi, Benkeser, Díaz, and van der Laan <doi:10.48550/arXiv.2205.05777>).
Last updated 2 months ago
causal-inferenceconditional-density-estimatesdensity-estimationhighly-adaptive-lassoinverse-probability-weightsmachine-learningnonparametric-regressionpropensity-score
7.34 score 17 stars 3 packages 71 scripts 450 downloadsbiotmle - Targeted Learning with Moderated Statistics for Biomarker Discovery
Tools for differential expression biomarker discovery based on microarray and next-generation sequencing data that leverage efficient semiparametric estimators of the average treatment effect for variable importance analysis. Estimation and inference of the (marginal) average treatment effects of potential biomarkers are computed by targeted minimum loss-based estimation, with joint, stable inference constructed across all biomarkers using a generalization of moderated statistics for use with the estimated efficient influence function. The procedure accommodates the use of ensemble machine learning for the estimation of nuisance functions.
Last updated 24 days ago
regressiongeneexpressiondifferentialexpressionsequencingmicroarrayrnaseqimmunooncologybioconductorbioconductor-packagebioconductor-packagesbioinformaticsbiomarker-discoverybiostatisticscausal-inferencecomputational-biologymachine-learningstatisticstargeted-learning
5.20 score 4 stars 5 scripts 292 downloadstxshift - Efficient Estimation of the Causal Effects of Stochastic Interventions
Efficient estimation of the population-level causal effects of stochastic interventions on a continuous-valued exposure. Both one-step and targeted minimum loss estimators are implemented for the counterfactual mean value of an outcome of interest under an additive modified treatment policy, a stochastic intervention that may depend on the natural value of the exposure. To accommodate settings with outcome-dependent two-phase sampling, procedures incorporating inverse probability of censoring weighting are provided to facilitate the construction of inefficient and efficient one-step and targeted minimum loss estimators. The causal parameter and its estimation were first described by Díaz and van der Laan (2013) <doi:10.1111/j.1541-0420.2011.01685.x>, while the multiply robust estimation procedure and its application to data from two-phase sampling designs is detailed in NS Hejazi, MJ van der Laan, HE Janes, PB Gilbert, and DC Benkeser (2020) <doi:10.1111/biom.13375>. The software package implementation is described in NS Hejazi and DC Benkeser (2020) <doi:10.21105/joss.02447>. Estimation of nuisance parameters may be enhanced through the Super Learner ensemble model in 'sl3', available for download from GitHub using 'remotes::install_github("tlverse/sl3")'.
Last updated 2 months ago
causal-effectscausal-inferencecensored-datamachine-learningrobust-statisticsstatisticsstochastic-interventionsstochastic-treatment-regimestargeted-learningtreatment-effectsvariable-importance
5.12 score 14 stars 19 scripts 229 downloadsbiotmle - Targeted Learning with Moderated Statistics for Biomarker Discovery
Tools for differential expression biomarker discovery based on microarray and next-generation sequencing data that leverage efficient semiparametric estimators of the average treatment effect for variable importance analysis. Estimation and inference of the (marginal) average treatment effects of potential biomarkers are computed by targeted minimum loss-based estimation, with joint, stable inference constructed across all biomarkers using a generalization of moderated statistics for use with the estimated efficient influence function. The procedure accommodates the use of ensemble machine learning for the estimation of nuisance functions.
Last updated 3 years ago
regressiongeneexpressiondifferentialexpressionsequencingmicroarrayrnaseqimmunooncologybioconductorbioconductor-packagebioconductor-packagesbioinformaticsbiomarker-discoverybiostatisticscausal-inferencecomputational-biologymachine-learningstatisticstargeted-learning
4.90 score 4 stars 5 scriptstmle3shift - Targeted Learning of the Causal Effects of Stochastic Interventions
Targeted maximum likelihood estimation (TMLE) of population-level causal effects under stochastic treatment regimes and related nonparametric variable importance analyses. Tools are provided for TML estimation of the counterfactual mean under a stochastic intervention characterized as a modified treatment policy, such as treatment policies that shift the natural value of the exposure. The causal parameter and estimation were described in Díaz and van der Laan (2013) <doi:10.1111/j.1541-0420.2011.01685.x> and an improved estimation approach was given by Díaz and van der Laan (2018) <doi:10.1007/978-3-319-65304-4_14>.
Last updated 2 months ago
causal-inferencemachine-learningmarginal-structural-modelsstochastic-interventionstargeted-learningtreatment-effectsvariable-importance
4.83 score 16 stars 42 scriptsmedoutcon - Efficient Natural and Interventional Causal Mediation Analysis
Efficient estimators of interventional (in)direct effects in the presence of mediator-outcome confounding affected by exposure. The effects estimated allow for the impact of the exposure on the outcome through a direct path to be disentangled from that through mediators, even in the presence of intermediate confounders that complicate such a relationship. Currently supported are non-parametric efficient one-step and targeted minimum loss estimators based on the formulation of Díaz, Hejazi, Rudolph, and van der Laan (2020) <doi:10.1093/biomet/asaa085>. Support for efficient estimation of the natural (in)direct effects is also provided, appropriate for settings in which intermediate confounders are absent. The package also supports estimation of these effects when the mediators are measured using outcome-dependent two-phase sampling designs (e.g., case-cohort).
Last updated 9 months ago
causal-inferencecausal-machine-learninginverse-probability-weightsmachine-learningmediation-analysisstochastic-interventionstargeted-learningtreatment-effects
4.34 score 13 stars 17 scriptsmedshift - Causal mediation analysis for stochastic interventions
Estimators of a parameter arising in the decomposition of the population intervention (in)direct effect of stochastic interventions in causal mediation analysis, including efficient one-step, targeted minimum loss (TML), re-weighting (IPW), and substitution estimators. The parameter estimated constitutes a part of each of the population intervention (in)direct effects. These estimators may be used in assessing population intervention (in)direct effects under stochastic treatment regimes, including incremental propensity score interventions and modified treatment policies. The methodology was first discussed by I Díaz and NS Hejazi (2020) <doi:10.1111/rssb.12362>.
Last updated 3 years ago
causal-inferenceinverse-probability-weightsmachine-learningmediation-analysisstochastic-interventionstargeted-learningtreatment-effects
3.69 score 9 stars 11 scriptsnima - Nima Hejazi's R Toolbox
Miscellaneous R functions developed as collateral damage over the course of work in statistical and scientific computing for research. These include, for example, utilities that supplement existing idiosyncrasies of the R language, extend existing plotting functionality and aesthetics, help prepare data objects for imputation, and extend access to command line tools and systems-level information.
Last updated 5 years ago
3.22 score 33 scripts 152 downloadstmle3mediate - Targeted Learning for Causal Mediation Analysis
Targeted maximum likelihood (TML) estimation of population-level causal effects in mediation analysis. The causal effects are defined by joint static or stochastic interventions applied to the exposure and the mediator. Targeted doubly robust estimators are provided for the classical natural direct and indirect effects, as well as the more recently developed population intervention direct and indirect effects.
Last updated 3 years ago
causal-inferencecausal-mediation-analysismachine-learningmediation-analysisstochastic-interventionstargeted-learningtreatment-effects
2.68 score 3 stars 16 scripts