how to read the outcome of the test_prgds.py

[xnfei]

I tried python tests/test_prgds.py, but the result is a bunch of numbers
May I ask what does the outcome mean?

Using numpy backend.
seed: 7411
29% missing
0
500
1000
1500
2000
2500
3000
3500
4000
4500
0
500
1000
1500
2000
2500
3000
3500
4000
4500

[Aaron Schein]

All of the tests implement a variation of Geweke testing (see Roger Grosse's excellent tutorial on Geweke testing). All of the test scripts for testing models (e.g., test_prgds.py) will run this alternative Geweke test and then serialize a P-P plot for each latent variable. If the test passes, then all of the P-P plots should be straight lines (see the attached example).

I've added a bash script that 1) cleans up the pp_plot/ directory , 2) compiles the code (if necessary), 3) runs the test, and then 4) opens up all the P-P plots. To run a test, cd to the src/ directory and then type:

bash tests/run_test.sh test_prgds.py

[xnfei]

thanks. Last quesiton: what data did you use in test_prgds.py?
I check the code but haven't found the answer.

[Aaron Schein]

This test generates synthetic data from the prior and then fits the model to the synthetic data. There is no external source of data.

[xnfei]

Thank you Aaron, for everthing!