Econ 570: Assignment 2
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
Econ 570: Assignment 2
We have discussed causal inference in experimental settings and the role of confounders and selection bias. This assignment is about exploring how different estimators perform under different Data Generating Processes (DGPs), and how including or not including the right exogenous variables affects the estimate of the treatment effect.
Pick at least 3 out of the following 4 settings. For each of them:
● Simulate a DGP
● Illustrate your DGP with a DAG
● Show, using a Monte Carlo experiment with sample sizes N=100 and N=1000, what the bias, RMSE and size of your treatment effect estimate is in cases a and b.
● Give an example of a real-life situation that might be consistent with the DGP.
1. Simulate a DGP where the outcome of interest depends on a randomly assigned treatment and some observed covariates. How does your estimate of the treatment effect parameter compare in the following two cases
a. You do not control for any covariates
b. You control for all the covariates that affect the outcome
2. Simulate a DGP with a confounder (common cause)
a. You fail to control for the confounder
b. You do control for the confounder
3. Simulate a DGP with selection bias into the treatment (variable in between the path from the treatment to the outcome)
a. You control for the variable in between the path from cause to effect
b. You do not control for the variable in between the path from cause to effect
4. Simulate a DGP where the outcome variable is overrepresented at 0.
a. You estimate the treatment effect parameter using the Conditional-on-Positives (COP) framework
b. You estimate the treatment effect using the conventional method of comparing the outcomes of treated and untreated individuals.
Check your final work into a repository. The repository should contain.
1. Jupyter notebook
2. Simulated data
3. Extra, not required: a separate python file with helper functions that are imported into the notebook
To turn in your assignment, make sure your repository is public and submit the link to the repository here https://forms.gle/TyENfjppBQWNxYAi9
2022-04-12