hal9001: Scalable highly adaptive lasso regression in R

Hejazi, Nima S.; Coyle, Jeremy; Laan, Mark J. van der

doi:10.21105/joss.02526

Cited by 20 publications

(18 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…

is fixed as 0.5 in our experimental design since we did not allow the respondents to opt out when choosing between two hypothetical vaccines. The highly adaptive LASSO is implemented according to hal2009 ( Hejazi et al (2020) ) with the default hyperparameter values. We calculate the estimated score function

, where

is the average potential outcome, which is estimated by excluding the fold that includes respondent i .…”

Section: Estimation Strategymentioning

confidence: 99%

See 1 more Smart Citation

Determinants of COVID-19 vaccine preference: A survey study in Japan

Kawata

Nakabayashi

2021

SSM - Population Health

View full text Add to dashboard Cite

Objective Vaccination is a critical measure for containing the COVID-19 pandemic. We survey the determinants that affect the preference for COVID-19 vaccines in Japan, a vaccine hesitant nation. Setting and design : We conducted a randomized conjoint analysis survey of the preference for vaccines on the Internet by recruiting a nonprobability sample of 15,000 Japanese adults. The survey assigned 5 choice tasks to the respondents. In each task, the respondents evaluated 2 hypothetical COVID-19 vaccines and were asked which they would choose. The vaccine attributes included efficacy, major and minor adverse side effects, country of vaccine development and clinical trial, and vaccine type. Treatment The choice task asked the participants to select a vaccine from 2 hypothetical vaccines as an optional vaccine or select a vaccine as a mandatory vaccine with a probability of 0.5 for each. Results Compared to vaccines developed in China, vaccines developed domestically or in the US raised the choice probability by 131% and 96%, respectively. A domestic clinical trial increased the choice probability by 33%. An increase in efficacy from 50% to 90% increased the choice probability by 43%. A decrease in the risk of severe side effects from 1 per 10 thousand to 1 per 1 million increased the choice probability by 38%. The vaccine type was irrelevant. Making vaccination compulsory increased the choice probability of China- and Russia-developed vaccines by 6% and 4%, respectively, and increased that of taking a high-risk vaccine by 5% and a modestly effective (70%) vaccine by 4%. General vaccination hesitancy, political positions, demographic characteristics, education, and income were irrelevant. Conclusions A domestically developed vaccine with a domestic clinical trial could substantially increase the preference for a vaccine. Making vaccination compulsory could modestly reduce the penalty for a vaccine with side effects, geopolitical, and efficacy concerns.

show abstract

“…

, where

is the average potential outcome, which is estimated by excluding the fold that includes respondent i .…”

Section: Estimation Strategymentioning

confidence: 99%

“…

Section: Estimation Strategymentioning

confidence: 99%

Determinants of COVID-19 vaccine preference: A survey study in Japan

Kawata

Nakabayashi

2021

SSM - Population Health

View full text Add to dashboard Cite

show abstract

“…Fit the HAL with 1 -norm, obtain the set of basis functions and a starting value of , denote it as . This is a CV-optimal value of the penalty parameter returned by the hal9001 package 58,59 , which is essentially from the "cv.glmnet" function with 10-fold cross-validation.…”

Section: Undersmoothed Highly Adaptive Lassomentioning

confidence: 99%

Evaluating the Robustness of Targeted Maximum Likelihood Estimators via Realistic Simulations in Nutrition Intervention Trials

Li¹,

Rosete²,

Coyle³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Several recently developed methods have the potential to harness machine learning in the pursuit of target quantities inspired by causal inference, including inverse weighting, doubly robust estimating equations and substitution estimators like targeted maximum likelihood estimation. There are even more recent augmentations of these procedures that can increase robustness, by adding a layer of cross-validation (cross-validated targeted maximum likelihood estimation and double machine learning, as applied to substitution and estimating equation approaches, respectively). While these methods have been evaluated individually on simulated and experimental data sets, a comprehensive analysis of their performance across "real-world" simulations have yet to be conducted. In this work, we benchmark multiple widely used methods for estimation of the average treatment effect using ten different nutrition intervention studies data. A realistic set of simulations, based on a novel method, highly adaptive lasso, for estimating the data-generating distribution that guarantees a certain level of complexity (undersmoothing) is used to better mimic the complexity of the true data-generating distribution. We have applied this novel method for estimating the data-generating distribution by individual study and to subsequently use these fits to simulate data and estimate treatment effects parameters as well as their standard errors and resulting confidence intervals.Based on the analytic results, a general recommendation is put forth for use of the cross-validated variants of both substitution and estimating equation estimators. We conclude that the additional layer of cross-validation helps in avoiding unintentional over-fitting of nuisance parameter functionals and leads to more robust inferences.

show abstract

“…This can generally be implemented with LASSO software implementations such as glmnet in R [4]. Additionally, HAL9001 provides implementations for linear, logistic, Cox, and Poisson regression, which also provides HAL estimators of the conditional hazards and intensities [7].…”

Section: Definition Of Hal-mlementioning

confidence: 99%