Off by 1 error in negbin update

[cErikson]

There appears to a off by one error in negbin update function.
Adding a -1 to the index seems to resolve the issue.
self.acceptances[self.sample_iter-1] = True

Traceback (most recent call last)
Cell In[48], line 1
----> 1 mcmc_negbin = md2.negbin.run_graph(
2 mcmc_negbin,
3 crash_if_error=True
4 )

File ~/miniforge3/envs/mdsine2/lib/python3.11/site-packages/mdsine2/negbin.py:1053, in run_graph(mcmc, crash_if_error, log_every)
1038 '''Run the MCMC chain mcmc. Initialize the MCMC chain with build_graph
1039
1040 Parameters
(...)
1050 mdsine2.BaseMCMC
1051 '''
1052 try:
-> 1053 mcmc.run(log_every=log_every)
1054 except Exception as e:
1055 logger.critical('CHAIN {} CRASHED'.format(mcmc.graph.name))

File ~/miniforge3/envs/mdsine2/lib/python3.11/site-packages/mdsine2/pylab/inference.py:524, in BaseMCMC.run(self, log_every, benchmarking)
522 try:
523 with time_counter() as t:
--> 524 node.update()
525 node.add_trace()
527 if benchmarking:

File ~/miniforge3/envs/mdsine2/lib/python3.11/site-packages/mdsine2/negbin.py:311, in NegBinDispersionParam.update(self)
309 u = np.log(pl.random.misc.fast_sample_standard_uniform())
310 if r > u:
--> 311 self.acceptances[self.sample_iter] = True
312 self.temp_acceptances += 1
313 else:

IndexError: index 2000 is out of bounds for axis 0 with size 2000

[yk23]

Hi cErikson,

This is a weird error that we've never encountered before. I'm having trouble reproducing it.

Under what settings did you run the infer-negbin command? Did you specify any burnin?

To my understanding, the sample_iter variable starts at zero so it should increment exactly up to len - 1.