Causal Data Fusion
Looking at an unknown object from more than one angle may help in gaining a better understanding of it. In many applications, due to the complexities in the setting, the information in the data is not sufficient for identification or efficient estimation of the causal parameter of interest. However, it is the case that we have access to data from our system of interest in other domains. Formally, we have data from more than one distribution from the same causal system. A natural question is how to combine the data from several domains for causal inference.
Here is a list of my contributions to the topic of causal data fusion:
Here is a list of my contributions to the topic of causal data fusion:
- Data fusion for identification and estimation of long-term causal effects [link]
- Data fusion for identification and estimation of parameters in the presence of non-ignorable missing data [link]
- A general identification algorithm for data fusion using graphical models [link]
- Causal discovery based on data fusion from observational and experimental domains [link] [link] [link]