New answers tagged causality
1
vote
How can you rewrite the estimand in terms of propensity scores? Dowhy question
The casual estimand is $E[Y|do(a)]$. When the backdoor criterion is satisfied, we have that $E[Y|do(a)] = E[E[Y|A=a, X]]$, where $X$ is a sufficient adjustment set. The propensity score theorem says ...
- 28.1k
1
vote
Difference-in-differences parallel trends for treatment and control groups of uneven size
One common way of supporting the parallel trends is using the relative time model. I'm assuming in your model, you have the treatments at different time periods (P1--P5), so first you need to have a ...
- 31
1
vote
Accepted
Synthetic control method based on several treated units
Yes, it would be fine. To quote Abadie (2021) Using Synthetic Controls:
Feasibility, Data Requirements, and Methodological Aspects: "The synthetic control framework can easily accommodate ...
- 40.5k
1
vote
Representing interaction effects in directed acyclic graphs
A new method of representing interactions by creating dedicated nodes was proposed and termed "IDAG" since this question was asked. In my understanding the example sentence from question &...
- 111
0
votes
Accepted
Understanding the Difference Between Independent and Dependent Variables
Variable Definitions
The wording conveys, with caution, the nature of each variable.
The dependent variable is the variable being acted upon. It is dependent on the influence of other factors. ...
- 4,170
0
votes
DAG - why is the path open?
As I understand it, the paths D <- Ed -> St -> P -> Su and D <- A -> P -> Su are both closed because the contain the collider P.
That's not quite right. I believe the key thing ...
- 945
0
votes
Accepted
DAG - why is the path open?
$P$ would only be a collider if you were examining a path such as $A\to P\leftarrow St.$ That is, whether a node is a collider or not is actually dependent on the path you take through that node.
So ...
- 5,605
0
votes
Noise abduction for computing counterfactuals
How can we obtain the noise variable values without knowing the structural equations?
The short answer is we cannot obtain them, at least not without further assumptions.
The paper on counterfactual ...
- 505
0
votes
Does causation imply correlation?
I add a less statistically technical answer here for the less statistically inclined audience:
One variable (let's say, X) can positively influence another variable (let's say, $Y$), while not being ...
- 191
1
vote
Accepted
How to understand the second rule of front door criterion?
I can understand your confusion. In the Front-Door Criteria of Definition 3.4.1, the original text for #2 read, "There is no unblocked path from $X$ to $Z$." The book's errata changed that ...
- 5,605
1
vote
How should control candidates be decided for causal inference?
The attrition issues you are facing here are common issues in trials involving a treatment imposed by the researchers. In many contexts this is dealt with by considering causal effects on the basis ...
- 112k
1
vote
How should control candidates be decided for causal inference?
The first question you have to ask yourself when you are facing a problem like this is "What exactly am I trying to estimate?", and be precise. This will usually start to guide you in the ...
- 21
1
vote
Accepted
Biased sample same like conditioning on a collider?
Not sure what you mean by a biased sample exactly, but we often have the issue of selection bias, which can result from conditioning on a common cause of the intervention and outcome (or generally a ...
- 337
1
vote
Accepted
How to handle intermittent missing visits in marginal structural model?
So, you won't want to only use the available visits as you describe.
Most work transforms the data into monotone missingness, as it is easier to work with. To do this, you would censor someone when ...
- 1,752
0
votes
Causal counterfactual inference model comparison
The answer to your question is no, the p-value or the proximity of the estimated effect to the null hypothesis (in this case, I assume it's zero or no effect) are not good metrics for predicting how ...
1
vote
Accepted
In regression, should we adjust for variables only associated with the independent or dependent variable?
The answer is that, since there is no connection with both the cause and the effect, it is impossible for age at scan to be a confounding variable. It cannot set up a backdoor path from the cause to ...
- 5,605
1
vote
Accepted
Bang and Robins doubly-robust estimator biased and with large variance?
This is a slight error in how you programmed the DR estimator. Although dr_weight is the clever covariate, it is actually a function of the treatment. That means ...
- 28.1k
1
vote
Causal inference for intervention with no data in pre-intervention period
If treated stores are (truly) randomly selected, this is an experimental setup, and you can derive ATT straightforwardly in a number of ways. Random selection with an appropriate sample size does a ...
- 347
20
votes
Incorrectly Using the Word "Causal" to Describe a Regression Model?
There is a very careful formulation in Gelman, Hill, and Vehtari Regression and Other Stories:
From the data alone, a regression only tells us about comparisons between units, not about changes ...
- 3,932
19
votes
Incorrectly Using the Word "Causal" to Describe a Regression Model?
On average, a one unit increase in $x_i$ causes is associated with an increase in $y_i$ of $\beta_1$ units.
- 15.8k
2
votes
Accepted
Can controlling for a variable block the backdoor path opened by controlling for a collider?
The short answer is no - controlling for x1 in this case may mitigate bias, but does not eliminate it: the problem arises because x2 is directly caused by the outcome. You can see this in a simple ...
- 393
1
vote
Accepted
Invalid use of Propensity Score Matching?
If you're willing to assume this particular causal model, then you don't need to use propensity score matching, or any other backdoor path related method, because the effect of $X$ on $Z$ is not ...
- 1,547
2
votes
Can controlling for a variable block the backdoor path opened by controlling for a collider?
Ok after reading your comments and thinking about for a little bit I both think, that this hypothetical scenario is kind of interesting and that I have a solution for most cases:
When you fit a linear ...
- 922
0
votes
How should control candidates be decided for causal inference?
As Henry pointed out, if the four groups are defined and each individual be able to be clearly identified into one of the group?
Assume the groups are correctly defined, from the analysis viewpoint, ...
- 31
1
vote
Synthetic Control Method: Why average pre-treatment controls?
The main reason is to 1) prevent overfitting and 2) ease interpretation.
To understand 1), compare two cases; in one case we only use pre-period averages and in another case we use every individual ...
1
vote
How to Evaluate and Visualize the Positivity Assumption for the ATT
For estimating the ATT, overlap means that the control gorup "encloses" the treated group; that is, the support of the treated group is in the support of the control group. The relaxedness ...
- 28.1k
1
vote
Accepted
Linear regression assumptions for causal Inference
One assumption is that the regression model is correctly specified. This assumption implies that (1) treatment and outcome are linear and (2) confounding variables and outcome are linear if only main ...
- 1,752
Top 50 recent answers are included