Experimental Causal Designs Flashcards

You may prefer our related Brainscape-certified flashcards:
1
Q

What is Experimental Causal Designs?

A

Experimental causal designs involve comparing counterfactual states of the world and explicitly assigning units to treatment conditions to estimate causal effects.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What is Causal Inference?

A
  • A causal inquiry involves comparing counterfactual states of the world.
  • A data strategy is experimental when units are explicitly assigned to treatment conditions.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What is a Two Arm Randomised Experiment?

A
  • Subjects are randomly assigned to one of two conditions (treatment/control or potentially two different treatments).
  • Each unit has two potential outcomes: treated and untreated.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What are Cluster Randomised Experiments?

A

Entire groups (e.g., classrooms, localities, households) are assigned to treatment or control conditions.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What are some challenges to Cluster Randomised Experiments?

A

Higher Variance

Clustering is generally less desirable statistically (higher variance) but may be necessary

Bias from Uneven Cluster Sizes

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What are Subgroup Designs?

A

Understand differences in treatment effects between subgroups. It is not randomised. Difference in CATEs.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What are Factorial Experiments?

A

Experiments with multiple treatments (e.g., Treatment 1 with 2 levels, Treatment 2 with 2 levels).

Analyze how the effect of one treatment depends on the level of another treatment.Interaction effects can reveal complex relationships between treatments.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Advantages of Factorial Experiments?

A
  • Efficiency:Address multiple research questions simultaneously,potentially using a smaller total sample size compared to running separate experiments for each question.
  • Rich Information:Explore how different treatments interact and influence each other’s effects (e.g.,how fertilizer A affects plant growth depends on whether irrigation is used).
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What are Placebo Controlled Experiments?

A

A placebo-controlled experiment involves comparing a treatment group receiving the actual treatment with a control group receiving a placebo, typically used to assess the effectiveness of a treatment by ensuring any observed effects are not simply due to the placebo effect.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

What are some advantages and disadvantages to placebo controlled experiments?

A

Benefit:

  • Can increase precision over encouragement designs, particularly when the underlying compliance rate is low.

Drawback:

  • Over/under sampling of placebo and treatment groups can lead to increased variance and reduced power, similar to encouragement designs.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What are Stepped Wedge Experiments?

A
  • Treatment is rolled out sequentially over multiple time periods to groups of units (e.g., schools, hospitals).
  • Sometimes referred to as a “waitlist design.”
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What are some advantages or disadvantages?

A

Strengths:

  • More ethical than withholding treatment from a group entirely.
  • May have more power for the same number of total units compared to a two-arm randomized experiment (due to sequential rollout and observing units over time).

Weaknesses:

  • More expensive than a two-arm randomized experiment for the same number of units.
  • Not ideal when time effects impact different units differently (heterogeneity in time effects).
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What are Randomised Saturation Experiments?

A

Multi-level design with cluster randomization, used when the treatment effect depends on the treatment status of other units within a hierarchical structure (e.g., vaccination rates within a neighborhood). Analyzes how the saturation level (proportion of treated units within a cluster) influences treatment effects.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly