Jump to content

Debugging Experiments


MEagar

Recommended Posts

Good Day,


I have tried to search the blog yet came up short. I am sure the capability exists, just not sure where to look for it.


My normal model works fine, yet my experiments fail. This is not an error i.t.o. defining the experiments but its somewhere in my code. This is predominately brought on by decision logic based on "Run.mode ==2" assignments.


Anybody knows how I can find a trace window or some sort of indication where it is failing in the experiment ??


FANKS

Link to comment
Share on other sites

Hi All,


Thanks to Christine for assisting me on this and I would thus like to share the methodology on the Blog for it may save you some time.


When setting up Experiments, you define controls. These controls can be assigned in the process logic with a simple decide Block with a specified condition "Run.Mode == 2".


When true, it will use the experimenter controls as inputs and when false, it will use the interactive inputs. See attached. This becomes VERY useful when using Excel Tables as inputs, which the Experimenter needs to overwrite.


As the Experimenter does not have a trace window, by defaulting the decide block to true, and using the Controls section of the model property, see attached, as specified inputs, you impersonate the interactive run as a scenario being executed.


All errors thrown in the Experimenter, which is not visible during the interactive run, will become evident.


I hope this helped and saved some people some time. Was pulling my hair out for the experimenter had picked up errors which was not detected during the interactive mode.


Mark

516257400_ControlProperties.PNG.a5d623f006b470733a81784d98a9357f.PNG

Run_Mode.PNG.72eeb195f2a44edbad15ad5c1ab3f0b3.PNG

Link to comment
Share on other sites

Thanks Mark, very useful information.


RE:

All errors thrown in the Experimenter, which is not visible during the interactive run
FYI, the problem with errors not being reported in the experiment window was a bug that has been fixed in 6.96.
Link to comment
Share on other sites

  • 2 weeks later...

Hi Mark


I am having exactly the same problem as you were - no errors in interactive mode, but two replications out of five fail in an experiment. Try as I might in interactive mode, I cannot raise an error. I have read your post and tried this out, but it doesn't change anything. I still get no errors in interactive mode, but two failed replications out of five in an experiment. The replications are seven days each, I have run in interactive mode for 10 days with nothing. Is it just a case of running for longer until the error (assuming it is one in both cases) happens? Any ideas?

Link to comment
Share on other sites

Cstein,

Are you using 5.96 or later so that the experiment reports error messages as it should? If not, updating may resolve your problem.


In interactive mode:

  • Ensure that you are using exactly the same values for ALL controls (aka referenced properties).
  • By default, interactive mode only creates replication #1. If you want interactive mode can run a different replication. If you are getting an error in replication 3 for example, change the replication number to 3 in Run Ribbon -> Advanced Options -> Randomness -> Replication Number.

Link to comment
Share on other sites

  • 2 weeks later...

Thank you for the tips, I will try them. Embarrassingly though, the problem at the time that I posted the last message turned out to be the fact that I was writing values to a CSV file with a write step and had opened two of the files to look at the values, and had forgotten to close them. So those two replications kept failing because the data could not be written. :oops:


-Chris

Link to comment
Share on other sites

×
×
  • Create New...