Cloak and DAG: a response to the comments on our comment

Area of Science:

• Neuroimaging
• Causal Inference
• Statistical Modeling

Background:

• Directed graphical models (DGMs), including structural equation models (SEMs), are used in neuroimaging to estimate causal effects.
• Researchers may be unaware of the underlying assumptions required for valid causal interpretation of these models.

Purpose of the Study:

• To clarify the assumptions neuroimaging researchers make when using DGMs and SEMs for causal inference.
• To emphasize the importance of evaluating these assumptions before interpreting model parameters as causal effects.

Main Methods:

• The study critically examines the assumptions inherent in applying DGMs and SEMs to neuroimaging data.
• It discusses the implications of unmet assumptions for causal effect estimation.

Main Results:

• Parameters from DGMs/SEMs should not be interpreted as causal effects if underlying assumptions are violated.
• Alternative methods or experimental designs may be necessary when assumptions are unrealistic.

Conclusions:

• Neuroimaging researchers must carefully consider the plausibility of DGM/SEM assumptions for their specific research questions, particularly concerning causation and effective connectivity.
• Awareness and validation of these assumptions are crucial for reliable causal inference in neuroimaging.