Advanced Research Methods
Knowledge video; Directed Acyclic Graphs
Learning goals:
After watching this knowledge video you will be able to:
- Understand DAG terminology
- Apply DAG rules to answer a research question
> You will be able to determine the effect of one variable to another, without the influence of
other disruptive variables.
DAG theory;
Directed Acyclic Graphs (DAGs) are graphical representations of the causal structure underlying a
research question.
Going to the gym (exposure X) → Health (outcome Y).
But you also need to know: diet, lifestyle and disease. This can affect the relation between X and Y. In
order to examine the relation unbiased, you need to account to this matter. You have to remove the
disruptive variables.
- DAGs help to visualize the causal structure underlying a research question
- You need a priori theoretical / subject knowledge about the causal structure to draw a DAG
(e.g., from previous studies, literature, common sense).
- Collect data on all relevant variables
- ‘Simple’ rules can be applied to determine for which variables to adjust in regression analysis
and how to interpret the results.
DAG terminology
1. Paths
2. Causal paths and backdoor paths
3. Open and closed paths, and colliders
4. Blocking open paths
5. Opening blocked paths
1. Paths
RQ: Influence of X on Y?
• A path is any route between exposure X and outcome Y. They always relate directly to the
research question. They follow any route between X and Y.
• Paths do not have to follow the direction of the arrows.
Q: How many paths are there between exposure X and outcome Y?
A: 4 paths from X and Y
,In this DAG the structure that underlies the research question, is described by 4 paths. We are only
interest in the relation between X and Y, having data of all the paths is important to describe the
relation in an unbiased matter.
2. Causal paths and backdoor paths
RQ: Influence of X on Y?
• A causal path follows the direction of the arrows
• A backdoor path does not
Q: Which are causal or backdoor paths?
A: 2 causal paths (X -> Y) (X -> V -> Y)
2 backdoor paths (X <- L -> Y) (X -> W <- Y)
3. Open and closed paths
RQ: Influence of X on Y?
Regardless of paths are causal or backdoor.
• All paths are open, unless they collide somewhere on a path (in a variable, on a path between
X and Y).
• A path is closed if arrows collide in one variable on that path (that variable is called a collider)
Q: How many of the four paths that we have identified are open and how many are closed?
A: 3 open paths, 1 closed path
4. Blocking open paths
RQ: Influence of X on Y?
, • Open (causal or backdoor) paths transmit association (association = relationship)
• The association between X and Y consists of the combination of all open paths between them
• Here: all paths except X -> W <- Y (because W is a collider)
To examine the direct association between X and Y, we need to remove all variables that are not
directly relevant and only disruptive, by blocking those open paths.
• An open path is blocked when we adjust for a variable (L) along the way
• This means that we remove the disruptive influence of L from the association between X and
Y
• How? By including variable L in the regression analysis
• Backdoor paths always need to be closed
• Causal paths need to be open/ closed depending on RQ
5. Open blocked paths
RQ: Influence of X on Y?
• Including a collider (W) in the analysis means you open the blocked backdoor path. This will
disrupt the association.
• This introduces bias in the association between X and Y. To obtain a unbiased effect, you need
to close the backdoor path. For example by removing a collider, or you can include another
variable that can close the path again.
Key points;
After watching this knowledge video you are expected to:
- Understand DAG terminology
- Be able to apply DAG rules to answer a research question in an unbiased manner (block
backdoor paths!).
, Lecture 1.2; Introduction to causal inference
Learning goals;
After lecture 1.2 you will be able to:
- Explain the potential outcomes approach in causal inference
- Apply the approach in critical thinking about causal effect estimation
- Define ‘causal effect’
- Apply the concepts of consistency, positivity, and exchangeability in randomised and
observation trials
Causal claims are everywhere (papers. magazine etc. )
Is the powder really that good?
‘Improves the quality of you skin’ implies a causal effect
- A leads to B
- Use of the True Match Minerals powder leads to a better skin
Question: Would you buy the powder?
- Is the scientific evidence convincing?
- What are arguments for and against buying the powder?
➢ Small sample size
> Is this always a problem? Depends on research question
➢ Study performed or financed by commercial company
> Is this always a problem?
➢ No control group: very important
> Essential data missing
> What would happen without treatment?
> Potential regression to the mean
What do we want to know?
In causal inference:
- We are not interested in the outcome per se (i.e. 70% less imperfection), but…
- We are interested in the role of the treatment in achieving this outcome (i.e. without True
Match Minerals powder, would there have been less skin imperfections?)
Conclusion:
We do not have that information
No causal claim can be made based on the study of Oreal
Causal effect
Formal definition by Hernan and Robins;
‘In an individual, a treatment has a causal effect if the outcome under treatment 1 would be different
from the outcome under treatment 2’
To asses this, we need information on:
➔ What would have happened?
➔ What will happen?
Assume that we know what would have happened in the Oreal study:
- Women A treated with True Match Minerals powder: 2 bad sports
- Had women A not been treated with True Match Minerals: 5 bad spots
➔ Individual treatment effect: -3 spots (or 60% less imperfections)
➔ Average treatment effect: average of individual effects in a population
Knowledge video; Directed Acyclic Graphs
Learning goals:
After watching this knowledge video you will be able to:
- Understand DAG terminology
- Apply DAG rules to answer a research question
> You will be able to determine the effect of one variable to another, without the influence of
other disruptive variables.
DAG theory;
Directed Acyclic Graphs (DAGs) are graphical representations of the causal structure underlying a
research question.
Going to the gym (exposure X) → Health (outcome Y).
But you also need to know: diet, lifestyle and disease. This can affect the relation between X and Y. In
order to examine the relation unbiased, you need to account to this matter. You have to remove the
disruptive variables.
- DAGs help to visualize the causal structure underlying a research question
- You need a priori theoretical / subject knowledge about the causal structure to draw a DAG
(e.g., from previous studies, literature, common sense).
- Collect data on all relevant variables
- ‘Simple’ rules can be applied to determine for which variables to adjust in regression analysis
and how to interpret the results.
DAG terminology
1. Paths
2. Causal paths and backdoor paths
3. Open and closed paths, and colliders
4. Blocking open paths
5. Opening blocked paths
1. Paths
RQ: Influence of X on Y?
• A path is any route between exposure X and outcome Y. They always relate directly to the
research question. They follow any route between X and Y.
• Paths do not have to follow the direction of the arrows.
Q: How many paths are there between exposure X and outcome Y?
A: 4 paths from X and Y
,In this DAG the structure that underlies the research question, is described by 4 paths. We are only
interest in the relation between X and Y, having data of all the paths is important to describe the
relation in an unbiased matter.
2. Causal paths and backdoor paths
RQ: Influence of X on Y?
• A causal path follows the direction of the arrows
• A backdoor path does not
Q: Which are causal or backdoor paths?
A: 2 causal paths (X -> Y) (X -> V -> Y)
2 backdoor paths (X <- L -> Y) (X -> W <- Y)
3. Open and closed paths
RQ: Influence of X on Y?
Regardless of paths are causal or backdoor.
• All paths are open, unless they collide somewhere on a path (in a variable, on a path between
X and Y).
• A path is closed if arrows collide in one variable on that path (that variable is called a collider)
Q: How many of the four paths that we have identified are open and how many are closed?
A: 3 open paths, 1 closed path
4. Blocking open paths
RQ: Influence of X on Y?
, • Open (causal or backdoor) paths transmit association (association = relationship)
• The association between X and Y consists of the combination of all open paths between them
• Here: all paths except X -> W <- Y (because W is a collider)
To examine the direct association between X and Y, we need to remove all variables that are not
directly relevant and only disruptive, by blocking those open paths.
• An open path is blocked when we adjust for a variable (L) along the way
• This means that we remove the disruptive influence of L from the association between X and
Y
• How? By including variable L in the regression analysis
• Backdoor paths always need to be closed
• Causal paths need to be open/ closed depending on RQ
5. Open blocked paths
RQ: Influence of X on Y?
• Including a collider (W) in the analysis means you open the blocked backdoor path. This will
disrupt the association.
• This introduces bias in the association between X and Y. To obtain a unbiased effect, you need
to close the backdoor path. For example by removing a collider, or you can include another
variable that can close the path again.
Key points;
After watching this knowledge video you are expected to:
- Understand DAG terminology
- Be able to apply DAG rules to answer a research question in an unbiased manner (block
backdoor paths!).
, Lecture 1.2; Introduction to causal inference
Learning goals;
After lecture 1.2 you will be able to:
- Explain the potential outcomes approach in causal inference
- Apply the approach in critical thinking about causal effect estimation
- Define ‘causal effect’
- Apply the concepts of consistency, positivity, and exchangeability in randomised and
observation trials
Causal claims are everywhere (papers. magazine etc. )
Is the powder really that good?
‘Improves the quality of you skin’ implies a causal effect
- A leads to B
- Use of the True Match Minerals powder leads to a better skin
Question: Would you buy the powder?
- Is the scientific evidence convincing?
- What are arguments for and against buying the powder?
➢ Small sample size
> Is this always a problem? Depends on research question
➢ Study performed or financed by commercial company
> Is this always a problem?
➢ No control group: very important
> Essential data missing
> What would happen without treatment?
> Potential regression to the mean
What do we want to know?
In causal inference:
- We are not interested in the outcome per se (i.e. 70% less imperfection), but…
- We are interested in the role of the treatment in achieving this outcome (i.e. without True
Match Minerals powder, would there have been less skin imperfections?)
Conclusion:
We do not have that information
No causal claim can be made based on the study of Oreal
Causal effect
Formal definition by Hernan and Robins;
‘In an individual, a treatment has a causal effect if the outcome under treatment 1 would be different
from the outcome under treatment 2’
To asses this, we need information on:
➔ What would have happened?
➔ What will happen?
Assume that we know what would have happened in the Oreal study:
- Women A treated with True Match Minerals powder: 2 bad sports
- Had women A not been treated with True Match Minerals: 5 bad spots
➔ Individual treatment effect: -3 spots (or 60% less imperfections)
➔ Average treatment effect: average of individual effects in a population
