Exploring Causal Pathways in Suicide Prevention
In my Experimental Design and Causal Inference class, under the supervision of Professor Valerie Nazzaro, I am conducting an independent analysis using the National Longitudinal Study of Adolescent Health (Add Health) dataset to explore the potential causal relationship between counseling and suicide attempts. The project aims to identify causal pathways leading to suicide attempts, focusing on whether counseling serves as a protective factor. To achieve this, I am using covariate matching to control for systematic differences between treatment and control groups, allowing me to isolate the effect of counseling and more accurately assess its role in preventing suicide attempts.
I also gained valuable skills in research design and data analysis throughout this course, which are outlined below:
Key Topics and Concepts Learned in Experimental Design and Causal Inference Class:
- Experimental Design Foundations: Principles of experimental design, including replication, randomization, balance, and blocking, essential for controlling confounding variables and establishing causality.
- Hypothesis Testing Techniques: In-depth understanding of statistical tests such as t-tests, ANOVA, and ANCOVA for evaluating differences between groups and testing causal hypotheses.
- Completely Randomized Design: Learning the general model structure of completely randomized designs and how to perform one-way ANOVA and multiple comparisons effectively.
- Randomized Block Designs: Extending experimental models to account for block effects and the benefits of blocking in controlling for confounding variables.
- Factorial Designs: Gaining familiarity with factorial design structures, assumptions, and methods for analyzing interaction effects among multiple factors, as well as addressing confounding.
- Quasi-Experimental Design: Exploring designs without control groups, such as repeated/removed treatment designs and case-control studies, and their relevance in observational data.
- Difference-in-Differences Models: Learning how to apply this technique for causal inference when dealing with longitudinal data, especially in policy or treatment evaluations.
- Regression Discontinuity Designs: Understanding the theory and practical applications of regression discontinuity designs, including potential threats to validity and how to address them.
- Multivariate Matching Techniques: Mastering advanced matching methods like propensity score matching and optimal pair matching to compare treated and control groups while accounting for multiple covariates.
- Causal Inference Methodologies: Building knowledge in identifying causal exposure through graph theory and matching estimators, and estimating causal effects without complete models using instrumental variable estimators.
- Statistical Software Proficiency: Gaining hands-on experience with R and RStudio for computational tasks, such as data visualization, hypothesis testing, and causal effect estimation.
- Ethical Considerations: Understanding the ethical challenges when conducting research in sensitive areas, particularly in health and social sciences, and ensuring the responsible handling of data.
