Help
Options
Didn't know it?
click below
click below
Knew it?
click below
click below
Don't Know
Remaining cards (0)
Know
retry
shuffle
restart
0:00
Embed Code - If you would like this activity on your web page, copy the script below and paste it into your web page.
Normal Size Small Size show me how
Normal Size Small Size show me how
MKTG 6600 Final
Final Exam flash cards
| Question | Answer |
|---|---|
| What are potential outcomes in causal inference? | Theoretical outcomes that could occur under each treatment conditions for a unit, defined before treatment assignment |
| Why can we never observe the individual treatment effect (ITE)? | Because we only observe one outcome per unit--the other is counterfactual and unobservable |
| What is the fundamental problem of causal inference? | We cannot observe both potential outcomes for the same unit, make causal effects unidentifiable without assumptions |
| Define exchangeability | Treatment groups are comparable in potential outcomes, conditional on covariates--no systematic differences due to confounding |
| What is a confounder? | A variable that affects both treatment assignment and the outcome, potentially biasing causal estimates |
| What does randomization achieve in causal inference? | It breaks the link between confounders and treatment, ensuring exchangeability and unbiased estimates |
| What is a balance table used for? | To compare covariate distributions across treatment groups and assess whether matching or weighting achieved balance |
| What is model dependency in matching? | When causal estimates vary significantly depending on the model specification, often due to poor overlap or imbalance |
| What is a propensity score? | The probability of receiving treatment given covariates, used to match or weight units to reduce confounding |
| Name four types of matching methods | Exact matching, coarsened exact matching (CEM), inverse probability weighting (IPW), full matching |
| What is the fundamental problem of causal inference in observational studies? (quiz 6) | Treatment effects on an individual level cannot be observed Core issue in causal inference is that one cannot observe both potential outcomes (w and w/out treatment) for the same individual |
| What is the main limitation of matching only observed characteristics? (quiz 6) | It cannot balance on unobserved characteristics While matching can control for observed differences, it does not account for unobserved variables that would also influence the treatment effect |
| What does matching in observational studies try to approximate? (quiz 6) | The balance achieved through randomization |
| How does matching prior to regression analysis enhance causal inference in observational studies? (quiz 6) | It makes treatment effect estimation less dependent on a model |
| What is the advantage of using balance tables in observational studies? (quiz 6) | They offer a method to compare treatment and control groups |
| Why might regression alone be inadequate for estimating causal effects in observational data? (quiz 6) | It mixes design and analysis phases |
| In the context of observational studies, what does the term "exchangeability" mean? (quiz 6) | Treatment and control groups are interchangeable |
| How does the use of IPW help in propensity score matching? (quiz 6) | By assigning weights to units |
| What is the primary goal when using observational data for causal inference? (quiz 6) | To mimic the structure of a randomized experiment |
| Why is it important to achieve balance in covariates between treatment and control groups in observational studies? (quiz 6) | To mimic the conditions of a randomized trial |
| Why is randomization considered the gold standard in experimental design? (quiz 6) | It guarantees that characteristics are balanced between the groups |
| What problem does coarsened exact matching (CEM) address? (quiz 6) | The difficulty in finding exact matches in high-dimensional data |
| Which of the following is a true statement about meta-learners in causal inference? (quiz 7) | They estimate heterogeneous treatment effects |
| What role does the propensity score play in causal inference using observational data? (quiz 7) | It estimates the treatment probability, given covariates |
| What is the advantage of using machine learning for causal inference over traditional statistical methods? (quiz 7) | It provides automatic model discovery |
| What does the S-learner specifically do to estimate treatment effects? (quiz 7) | It includes treatment assignment in a single model |
| What unique approach do casual forests use to estimate treatment effects? (quiz 7) | They modify the splitting criteria in tree algorithms |
| Which of the following best describes causal forests? (quiz 7) | A variation of random forests to estimate heterogeneous treatment effects |
| Why are causal forests considered to provide "doubly robust" estimates? (quiz 7) | They provide unbiased estimates if either the propensity score or the outcome model is correctly specified |
| What does conditional average treatment effect (CATE) measure? (quiz 7) | The variation in treatment effects across groups |
Created by:
Flutterpluff