Crash when marking changepoints in renaissance data

[vext01]

It could be that the data is incomplete (in terms of pexecs), but I get the following crash:

$ ~/research/warmup_stats/bin/mark_changepoints_in_json renaissance_results_outliers_w200.json.bz2
Marking changepoints and classifications.
Expecting a steady state to be reached before the last 500 iterations.
Using a fixed bound of 0.001s.
Using R version 3.3.3 and changepoint library 2.3.1
Loading: renaissance_results_outliers_w200.json.bz2
/home/vext01/research/warmup_stats/bin/mark_changepoints_in_json:227: RuntimeWarning: divide by zero encountered in log
  pen_value=15.0*numpy.log(len(p_exec)))
/home/vext01/research/warmup_stats/work/pylibs/rpy2/rinterface/__init__.py:186: RRuntimeWarning: Error in multiple.meanvar.norm(data, mul.method = method, penalty, pen.value,  :
  Minimum segment legnth is too large to include a change in this data

  warnings.warn(x, RRuntimeWarning)
Traceback (most recent call last):
  File "/home/vext01/research/warmup_stats/bin/mark_changepoints_in_json", line 298, in <module>
    main(options.json_files[0], options.delta, options.steady_state)
  File "/home/vext01/research/warmup_stats/bin/mark_changepoints_in_json", line 200, in main
    segments = get_segments(cpt, delta, steady_state, p_exec, outliers)
  File "/home/vext01/research/warmup_stats/bin/mark_changepoints_in_json", line 227, in get_segments
    pen_value=15.0*numpy.log(len(p_exec)))
  File "/home/vext01/research/warmup_stats/work/pylibs/rpy2/robjects/functions.py", line 178, in __call__
    return super(SignatureTranslatedFunction, self).__call__(*args, **kwargs)
  File "/home/vext01/research/warmup_stats/work/pylibs/rpy2/robjects/functions.py", line 106, in __call__
    res = super(Function, self).__call__(*new_args, **new_kwargs)
rpy2.rinterface.RRuntimeError: Error in multiple.meanvar.norm(data, mul.method = method, penalty, pen.value,  :
  Minimum segment legnth is too large to include a change in this data

[ltratt]

I assume this means that you have N iterations and a window size of M where N should be > M, but is actually <= M. I might be very wrong.

[vext01]

Well, that's what I thought too, but I have 2000 samples and I'm using the default window size.

I'll investigate when I get a chance. Just wanted to raise the bug so it isn't forgotten.

[Gomezale]

Hi Everyone, I'm pretty new in R use and trying to solve similar use but using FlowAI. When I run flowAI automated (over 142 samples) i nmiddle of the samples I get this error: data<- flow_auto_qc(fs_raw_cd45)

Any suggestion how to solve it? Thanks a lot in advance

[vext01]

This happens when the size of your data is less than the window size.

In the case I reported above, this was due to a crashed process execution, which Krun represents as the empty list.

My suggestion is to check the size of your data.

[Gomezale]

Thanks for your quick support. May I kindly ask you how to check the size of my data? and how to change to avoid the error?

[vext01]

It looks like you are using the changepoint library from R(?).

This repo is a Python library that merely consumes the R library in a black box fashion. As such, I can't really help with your query, but you may have luck asking upstream?