Latest COVID-19 Information

News

Judith Lok

Associate Professor

Research areas: causality, counterfactuals, longitudinal data, observational studies, competing risks, survival analysis, HIV/AIDS, personalized medicine, mediation analysis, clinical trial design.

 

Office MCS 232
Phone 617-353-0921
jjlok@bu.edu
Website
Office Hours Fall 2021: W: 9:30-10:30am, R: 2-3pm (via Zoom)
CV Judith Lok

 

 

Research

Dr. Judith Lok’s research program mostly focuses on causal inference methods and survival analysis in HIV-AIDS related problems, mainly using observational data. E.g., Dr Judith Lok is working on mediation analysis. She has proposed a new definition of direct and indirect effects, “organic” direct and indirect effects, an intervention based approach which obviates the need to be able to “set” the mediator to a specific value. The assumptions for organic direct and indirect effects are implied by the usual assumptions to identify natural direct and indirect effects. Furthermore, with Dr. Donna Spiegelman, Dr. Daniel Nevo, and MS student Ante Bing, she is working on Learn-As-you-GO or LAGO clinical trial designs, where the composition of the intervention package changes over time as earlier-stage outcomes become available. In addition, together with Dr. Constantin Yiannoustos, Dr. Agnes Kiragga, Elisavet Syriopoulou, Dr. Ron Bosch, and PhD student Dustin Ribadeau she has estimated the CD4 count and adherence trajectory over time after ART initiation in a sub-Saharan treatment center. To estimate the CD4 count trajectory, she has developed Inverse Probability of Censoring Weighting when the data are Missing Not At Random, but when an “outreach” sample is available: data on a selection of the patients who dropped out. Together with Dr. Ron Bosch and Dr. Michael Hughes she has been working on the course of CD4 after treatment in the long run (7-8 years), using the ALLRT data. The main issue with the ALLRT data, as is the case for most HIV cohort studies, was considerable dropout. Together with Dr. Ron Bosch, Dr. Mireille Schnitzer, and clinicians she has worked on how immune activation after HAART treatment predicts subsequent AIDS-defining and non-AIDS-defining events as well as immune function, using the observational ALLRT data. To predict the AIDS-defining and non-AIDS-defining events, they have developed efficient and doubly robust estimation of a prediction model for events in the presence of informative censoring. Together with Dr. Victor DeGruttola and Dr. Shu Yang she has worked on estimating the effect of ART as a function of time since infection using the observational AIEDRP data. In the AIEDRP data, the main issue is confounding by indication: patients who are treated and patients who are not treated are not comparable. To estimate the effect of ART as a function of time since infection, they have introduced a time-dependent version of coarse Structural Nested Mean Models, developed optimal estimation, considered model fit, and developed a sensitivity analysis to the assumption of no unmeasured confounding. Together with Dr. Michael Hughes, Dr. Brian Sharkey, and Dr. Shu Yang she has worked on safety outcomes in HIV clinical trials, using a competing risks model estimated in the presence of informative censoring and considering both overall effects and personalized medicine.

Dr. Judith Lok has developed an intellectual and mathematical framework for Structural Nested Models, both for continuous-time measurements and for discrete-time measurements. In collaboration with Dr. Richard Gill, Dr. Aad van der Vaart and Dr. James Robins, she has provided a rigorous proof of the asymptotic properties, consistency and asymptotic normality, of the resulting estimators in discrete time, which was published in Statistica Neerlandica. She has also provided the first proofs of these asymptotic properties in continuous time. This work has been published in the Scandinavian Journal of Statistics, “Structural nested models and standard software: a mathematical foundation through partial likelihood” and in the Annals of Statistics, “Statistical modeling of causal effects in continuous time” and “Mimicking counterfactual outcomes to estimate causal effects”. These methods have been applied to estimate treatment effects in HIV infected subjects. Dr. Judith Lok is still working on continuous-time structural nested models.

Education

PhD, Free University of Amsterdam, 2001

Software

Causal organic indirect and direct effects: closer to Baron and Kenny, with a product method for binary mediators: https://arxiv.org/abs/1903.04697 .

bootstrap_macro_Y_M_C_continuous_posted

run_wk4_M_C_continuous_posted

bootstrap_macro_Y_M_C_binary_posted

run_wk4_CARNA_NNRTI_binary_posted

pvalue_macro_posted