"bestIndex > maxIndex. Returning None." error message should include more detail

[penguinaugustus]

Hi,Professor Daniels,

It seems that i bump into some unknown mistake after i run the model and tried to display the result with p.plotBestModelResults(plotInitialConditions=True,indices=range(10));

I didn't get any plot back. Instead, it returned with

maxLogLikelihoodName: Warning: Only 9 of 9 fits have been performed.
maxLogLikelihoodName: bestIndex > maxIndex.  Returning None.
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-14-204fc9c0e868> in <module>
      1 plt.figure(figsize=(20,2))
----> 2 a = p.plotBestModelResults(plotInitialConditions=True,indices=range(10));
      3 
      4 for i in range(len(a[0])):
      5     a[0][i].plot(RNAtheoreticalData[0], RNAtheoreticalData[1+i],linewidth=1)

~/SirIsaac/SirIsaac/fittingProblem.py in plotBestModelResults(self, modelName, maxIndex, verbose, **kwargs)
    750         m = self.getBestModel(modelName=modelName,maxIndex=maxIndex,
    751             verbose=verbose)
--> 752         return self.plotModelResults(m,**kwargs)
    753 
    754 

~/SirIsaac/SirIsaac/fittingProblem.py in plotModelResults(self, model, filename, indices, plotFittingData, outOfSampleData, **kwargs)
    711 
    712         # plot model results
--> 713         plots = m.plotResults(fittingData,indepParamsList,
    714             plotFittingData=plotFittingData,
    715             outOfSampleData=outData,**kwargs)

AttributeError: 'NoneType' object has no attribute 'plotResults'

<Figure size 1440x144 with 0 Axes>

It seems that i run

p.fitAll()
fittingProblem.save(p,'mRNAconcentrationS2.pkl')

successfully.

Could you give me any hint on what may go wrong and how to fix it?

Thank you for your time!

Sincerely,
Yu

A detailed code and error file is attached here:
mRNA-sigmoidal.pdf

[bcdaniels]

Sorry that this error message is not very clear. By default, SirIsaac does not return a "best" model until it has found 3 models with more parameters that have a lower Bayesian likelihood than the model with the maximum likelihood. The maxIndex argument controls this behavior (as you can read about in the documentation for the function maxLogLikelihoodName).

One way around this is to call plotBestModelResults with the keyword argument maxIndex=-1, which will plot the model with the maximum likelihood, without the check. Another way is to give it the keyword argument modelName with the name of the model you want to plot (e.g. modelName='Model 7'). The "best" thing to do is to increase the number of models to be fit until you find the peak in the Bayesian likelihood using the default maxIndex.

Hope that helps. Let me know if you have other questions.

[penguinaugustus]

That make sense to me. Thanks a lot!

[bcdaniels]

Keeping this as a "to do" bookmark to update the error message.