Final Task do(code)

[danaderp]

Here is the final list of activities for do(code):

• Please prepare the granularity datasets to be compatible with our data frames. Follow the same structure of data frames as LibEST.
• Although we do not need to run word2vec or doc2vec on the granularity datasets, it is important to generate a preprocessed version. There is a preprocessing notebook that you can employ for this task.
• Please generate a causal model where the treatment is the granularity variability. You can analyze performance based on the outcomes of TC Tracer (ICSE2020)
• Please explain the step-by-step of how to model and evaluate a causal model in a nbdev notebook.
• added 0.0.0_howto_causal.ipynb

[scheurich-sarah]

@danaderp By "Follow the same structure of data frames as LibEST" do you mean include the same fields? Also, this preprocessing notebook, are you referring to main/nbs/0.0.[0,1,3,6,]_mgmnt.prep. series?

[danaderp]

@scheurich-sarah Yes. Same fields and notebook. You need to create your own script to extract the info from the new sources though.

[scheurich-sarah]

@danaderp I couldn't figure out the pre-processing bit, so apologies, but the first 3 check marks aren't done. I read predicted and ground truth links files they provided into dfs and aggregated them across all the projects, but that's all. In the files I have read into data frames, I have the following issues: one file fails when decoding using utf-8, I've seen some strategies for fixing that, but I'm not sure if that is something that would normally be caught/fixed with the proper pre-processing routine.