Discussionwk10

no coding part

Part 1. In your own words, please explain how does an RDD work? (2-3 sentences)

RDD exploits the idea that there is a sharp cutoff or threshold which dictates whether subjects receive a treatment or not, assuming subjects on either side of this threshold are similar except for the treatment. By comparing outcomes just above and below the cutoff, RDD aims to isolate the treatment effect, offering insights similar to a randomized experiment where the assignment is nearly random at the cutoff. The strict enforcement of the cutoff ensures that any difference in outcomes can be attributed to the treatment effect.

1. Are you convinced RD design can give you causal estimates? (1 sentence)

Yes, when the assumption that units just around the cutoff are similar holds true, RDD can provide reliable causal estimates, especially with appropriate bandwidth selection to focus on those observations close to the threshold.

2. When is RDD valid? (2-3 sentences)

RDD is valid when the assignment variable is not perfectly controllable by the subjects, ensuring that those near the cutoff are not systematically different in ways that affect the outcome. It also requires that the cutoff be sharp and precisely defined, with no manipulation of the assignment variable around the threshold, so that the distinction between those who receive the treatment and those who do not is based solely on the cutoff criterion. Proper bandwidth selection is crucial to ensure that the observations being compared are similar in all respects except for the treatment status.

3. When can RD design be invalid? (2 sentences) HINT: See the first bullet of Section 1 of the RDD_overview.pdf reading, and skim the first 2 pages of Section 3 of the RDD_overview.pdf reading.

RDD becomes invalid if subjects can manipulate the assignment variable to ensure they receive the treatment, known as precise control, which would introduce systematic differences between those just above and below the cutoff. It is also problematic if there are other discontinuities around the cutoff that could confound the treatment effect or if the chosen bandwidth includes observations that are not comparable, leading to biased estimates.

Part 2. ARTICLE CRITIQUE / YOUR INSIGHTS

Calonico, S., Jawadekar, N., Kezios, K., & Zeki Al Hazzouri, A. (20241). Regression discontinuity design studies: A guide for health researchers. The BMJ

1. Find any RDD paper and evaluate the study design. HINT: You can have a look at Table 4 RDD_overview.pdf and find many relevant RDD papers. Tell us what is the outcome variable, what is the running variable, who gets the treatment and why does the setup make sense? (5 lines max)

The outcome variable in the study is the incidence of cardiovascular disease. The running variable is systolic blood pressure, with individuals segmented based on a cutoff value of 140 mm Hg. Treatment refers to individuals with a systolic blood pressure above 140 mm Hg receive antihypertensive drugs. This setup makes sense because systolic blood pressure is a clinically relevant and commonly used measure to determine the necessity for antihypertensive treatment. The sharp cutoff at 140 mm Hg aligns with established clinical guidelines for hypertension management.

2. Is it a sharp or fuzzy RDD? (1 sentence)

It is a sharp RDD, as the treatment assignment is strictly based on whether the systolic blood pressure crosses the 140 mm Hg threshold.

3. Would you have done anything differently if you were running the experiment?

If I were running the experiment, I might consider including additional covariates that could influence both the treatment decision and health outcomes, such as age, gender, and other cardiovascular risk factors. This inclusion could help in controlling for confounding variables that might affect the results. Additionally, I might implement a fuzzy RDD design to account for any real-world non-compliance with the treatment assignment, providing a more nuanced understanding of the treatment’s effects among those who actually follow the prescribed regimen.