Figure: Diagram illustrating the concept of confounding.
Ideas related to Causal Inference
- Confounding: A situation in which a variable influences both the dependent variable and independent variable, causing a spurious association. Properly addressing confounders is crucial for unbiased estimation of causal effects.
- Treatment effect estimation: The process of determining the causal effect of a treatment or intervention on an outcome. This involves comparing the outcomes between treated and untreated groups while accounting for potential confounders.
- Propensity score (PS): The probability of receiving a treatment given a set of observed covariates. Propensity scores are used to balance the distribution of covariates between treated and untreated groups, facilitating causal inference (Karim 2021; Karim et al. 2022; Guadagni et al. 2024; Simoneau et al. 2022).
Figure: Steps of Propensity Score Modelling.
- Admin data-related
- High-dimensional Propensity Score (hdPS): An extension of the propensity score method that involves selecting a large number of covariates, often from administrative data, to better control for confounding (Karim 2024).
- High-dimensional Disease Risk Score (hdDRS): Similar to hdPS, but focuses on predicting disease risk by incorporating a large number of covariates, often used in observational studies to adjust for confounding.
- Double robust methods: Statistical methods that provide valid estimates of treatment effects if either the model for the treatment assignment or the model for the outcome is correctly specified. This enhances the robustness of causal inferences (Frank and Karim 2024).
- Targeted Maximum Likelihood Estimation (TMLE): A double robust method that combines machine learning algorithms with statistical techniques to estimate causal effects, providing efficient and unbiased estimates.
Cross-fitting: A technique used to reduce overfitting in machine learning by partitioning the data into multiple folds and fitting models on each fold. This method is often used in conjunction with double robust methods to improve causal inference (Mondol and Karim, n.d.).
Longitudinal: Studies that collect data from the same subjects repeatedly over time, allowing for the analysis of changes and the assessment of causal relationships over time (Karim et al. 2017).
References
Frank, Hanna A, and Mohammad Ehsanul Karim. 2024. “Implementing TMLE in the Presence of a Continuous Outcome.” Research Methods in Medicine & Health Sciences 5 (1): 8–19.
Guadagni, Stefano, Marco Catarci, Francesco Masedu, Mohammad Ehsanul Karim, Marco Clementi, Giacomo Ruffo, Massimo Giuseppe Viola, et al. 2024. “Abdominal Drainage After Elective Colorectal Surgery: Propensity Score-Matched Retrospective Analysis of an Italian Cohort.” BJS Open 8 (1): zrad107.
Karim, Mohammad Ehsanul. 2021. “Understanding Propensity Score Matching.” https://ehsanx.github.io/psw/.
———. 2024. “High-Dimensional Propensity Score and Its Machine Learning Extensions in Residual Confounding Control.” The American Statistician, no. just-accepted: 1–38.
Karim, Mohammad Ehsanul, Fabio Pellegrini, Robert W Platt, Gabrielle Simoneau, Julie Rouette, and Carl de Moor. 2022. “The Use and Quality of Reporting of Propensity Score Methods in Multiple Sclerosis Literature: A Review.” Multiple Sclerosis Journal 28 (9): 1317–23.
Karim, Mohammad Ehsanul, John Petkau, Paul Gustafson, Helen Tremlett, and The Beams Study Group. 2017. “On the Application of Statistical Learning Approaches to Construct Inverse Probability Weights in Marginal Structural Cox Models: Hedging Against Weight-Model Misspecification.” Communications in Statistics-Simulation and Computation 46 (10): 7668–97.
Mondol, MH, and ME Karim. n.d. “Crossfit: An r Package to Apply Sample Splitting (Cross-Fit) to AIPW and TMLE in Causal Inference.” GitHub Repository.
Simoneau, Gabrielle, Fabio Pellegrini, Thomas PA Debray, Julie Rouette, Johanna Muñoz, Robert W Platt, John Petkau, et al. 2022. “Recommendations for the Use of Propensity Score Methods in Multiple Sclerosis Research.” Multiple Sclerosis Journal 28 (9): 1467–80.