Strengthening Causal Inference with Sensitivity Analyses: Using the Bayesian Parametric G-Formula
Understanding causal relationships in real-world data is challenging, particularly in observational studies, where confounding can introduce bias. If confounding is not properly accounted for, treatment effect estimates may be misleading, impacting decision-making. If confounding is not properly accounted for, estimates of the treatment effect may be misleading, impacting decision-making.
Therefore, correctly modelling confounding is essential for valid causal inferences. Two popular approaches to achieve this are inverse probability of treatment weighting (IPTW) and the parametric g-formula. Both methods assume confounding has been adequately modelled. However, what if it is not?
Historically, this is listed as a limitation. In recent years, sensitivity analyses have become increasingly popular to better understand this assumption. This webinar will focus on using the Bayesian parametric g-formula as a sensitivity analysis for measured confounding. Prior information about confounders can be incorporated into the model. This allows for a better understanding of the range of the plausible effects. Various scenarios can also be considered including if the confounder is a weak confounder or strong confounder.
This approach offers a practical solution for researchers to ensure the validity of findings.
Date/Time: Thursday, 6th March 2025, 8am PT | 11am ET | 4pm GMT
Speaker: Ryan Batten, Senior Statistician, Phastar
Webinar Highlights:
• How Bayesian methods improve sensitivity analyses
• Incorporating prior knowledge about confounding
• Practical applications for real-world data research
Join us for an insightful webinar exploring how the Bayesian Parametric G-Formula can be used as a sensitivity analysis tool to strengthen causal inference and improve the validity of findings.
Your Statistical Expert
Ryan Batten
Senior Statistician, Phastar
Ryan is a Senior Statistician at Phastar. He has over five years of experience working with real-world data. He has used various study designs and datasets to answer causal questions. These include electronic health records, claims data, registries and wearables. He is also a part-time PhD student in Clinical Epidemiology at Memorial University. Additionally, he holds a MSc in Clinical Epidemiology, and a BEng in Civil Engineering.
Visit the Knowledge Centre
Access our library of webinar recordings including:
- FSP Strategic Partnerships & Fuel Efficient Drug Development
- Study Design Considerations for Biotechs
- Use of Expert Prior Elicitation for Clinical Research
