Open Yihuahai opened 4 days ago
Hi Yihuahai,
thanks for your question!
This Repository provides a collection of several classes and methods I've used for my masterthesis. The methods provided in the dml_sim
submodule expects a data generating process as a callable. The dml_emb
submodule provides methods to generate low dimensional embeddings based on image and text as inputs and expect a pandas DataFrame. Please refer to the Docstrings written down in the class headers, e.g.
dataset (pandas.DataFrame): the input dataset.
.
These embeddings can be used as confounders in several settings according to the DoubleML package:
https://docs.doubleml.org/
For more information on double machine learning with multimodal confounders please refer to our latest paper:
https://arxiv.org/abs/2402.01785
Thanks and have a nice day! BR Jan
What should the format of this dataset be, and are there any instance datasets to refer to