Causal Discovery with Generalized Linear Models through Peeling Algorithms
Author Information
Author(s): Wang Minjie, Shen Xiaotong, Pan Wei
Primary Institution: Binghamton University, State University of New York
Hypothesis
Can a novel method using peeling algorithms improve causal discovery in the presence of unmeasured confounders?
Conclusion
The proposed GAMPI method effectively identifies causal relationships and valid instruments, outperforming existing methods in the presence of confounders.
Supporting Evidence
- The GAMPI method was shown to yield correct discovery of all parent-child relationships.
- Numerical experiments demonstrated GAMPI's superior performance over state-of-the-art methods.
- The method was applied to Alzheimer's disease data, revealing significant gene-to-gene and gene-to-disease relationships.
Takeaway
This study introduces a new way to find cause-and-effect relationships between variables, even when some important information is missing.
Methodology
The study uses generalized linear models and two peeling algorithms to identify causal relationships and valid instruments.
Potential Biases
Potential biases from unmeasured confounders may still affect results despite the proposed methods.
Limitations
The method may not generalize well to all types of data or causal structures.
Participant Demographics
The study involved 712 subjects from the Alzheimer's Disease Neuroimaging Initiative dataset, categorized into four groups based on cognitive status.
Statistical Information
P-Value
0.05
Statistical Significance
p<0.05
Want to read the original?
Access the complete publication on the publisher's website