Replies: 2 comments 1 reply
-
Hi @hhegefeld , could you provide us with a bit more info on the data and output, to get a better idea of what's going on?
Some things to try naively (but better to work with more information):
|
Beta Was this translation helpful? Give feedback.
1 reply
-
@hhegefeld did you have any more luck on recent version of HSSM? |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi, I'm new to using DDMs. I'm trying to model a dataset hierarchically with only the random effect of participant as a "null" model to test more complex models against. I used this command:
null_model = hssm.HSSM(data=ddm_data, hierarchical=True, prior_settings="safe", link_settings="log_logit")
on data that looks like:The sampling runs, albeit with errors outputted: "Sampling 4 chains for 1_000 tune and 3_000 draw iterations (4_000 + 12_000 draws total) took 2259 seconds.
There were 12000 divergences after tuning. Increase
target_accept
or reparameterize.The rhat statistic is larger than 1.01 for some parameters. This indicates problems during sampling. See https://arxiv.org/abs/1903.08008 for details
The effective sample size per chain is smaller than 100 for some parameters. A higher number is needed for reliable rhat and ess computation. See https://arxiv.org/abs/1903.08008 for details"
The problem occurs when I try to plot the traces:
az.plot_trace(null_model.traces)
. I get the error: "RuntimeWarning: invalid value encountered in dividehist_dens = hist / (hist.sum() * np.diff(bin_edges))" which seems to be caused by an error computing:
kernel_n = int(bw * 2 * np.pi)
("OverflowError: cannot convert float infinity to integer").After looking online and finding several relevant posts (here and here), it seemed that the error may be caused "when all draws for a chain were identical". I looked at the posterior of the model, and in fact, the subject standard deviation for z seems to do just that (for at least 3 of 4 chains).
Is this a fundamental error with my model or data, or is it likely caused by version incompatibilities, etc? Thank you in advance for any help!
Note: I'm running through a JupyterLab Notebook in a conda environment on a computing cluster, if that is helpful to know.
Beta Was this translation helpful? Give feedback.
All reactions