amlab-amsterdam / cevae Goto Github PK

Dear authors,

Thanks for providing the code for IHDP. Would you be able to provide the implementation on the other datasets used in your paper?

Hi,

I'm trying to replicate your results on the synthetic data experiment in section 4.2. I'd appreciate it a lot if you could help me with several questions.

1. Which package did you use for implementing LR1 and LR2? Is it tensorflow or scikit-learning? If you used tensorflow, did you implement it as softmax regression (with two-dimensional output) or the standard logistic regression?

2. How did you evaluate the model? Did you do train-validation-test split? If so, what is the size of the test data?

3. Did you use early stopping when training the model? What was the stopping criterion you used?

Thanks a lot!

Best,
Haozhu

Hi,

I just read your paper. I have two questions about the meaning of Within-sample and out-of-sample in Table 1 and Table 2.

1. I didn't see the definition of these two words in the article, could you help me to explain?
2. Are these two Settings reflected in your code?

Thanks!

Hi,

I notice that there are quite a few missing values in the twins dataset. May I ask how they are handled in the experiments in your paper? How do you impute them?

Thanks!

According to the the original paper (Hill 2011) there are 747 units (139 treated, 608 control). In the data folder there are 10 csv.

Which is the original csv?
Also, Why are there are 10 csv? Are they simulated?

Dear Authors,

Thanks for your inspiring work.

I'm wondering whether you could share more details about the pre-processing procedure for the TWINS dataset. Specifically, did you keep the raw values of the categorical variables or convert them into one-hot encoding? In addition, for the features 'brstate' and
'brstate_reg', did you use both or just one of them?

Thanks a lot!

Best,
Haozhu