Error while running MixtureSustain

[SebastianVol]

Hi,

I am currently using the Mixture Sustain and I get the following error:

``
pySuStaIn/AbstractSustain.py:556: RuntimeWarning: invalid value encountered in divide
total_prob_subtype_norm = total_prob_subtype / np.tile(np.sum(total_prob_subtype, 1).reshape(len(total_prob_subtype), 1), (1, N_S))
pySuStaIn/AbstractSustain.py:557: RuntimeWarning: invalid value encountered in divide
total_prob_stage_norm = total_prob_stage / np.tile(np.sum(total_prob_stage, 1).reshape(len(total_prob_stage), 1), (1, nStages + 1)) #removed total_prob_subtype
pySuStaIn/AbstractSustain.py:560: RuntimeWarning: invalid value encountered in divide
total_prob_subtype_stage_norm = total_prob_subtype_stage / np.tile(np.sum(np.sum(total_prob_subtype_stage, 1, keepdims=True), 2).reshape(nSamples, 1, 1),(1, nStages + 1, N_S))

---

ValueError Traceback (most recent call last)
Cell In[19], line 8
1 #************* run SuStaIn to infer subtype sequences and subjects' subtypes/stages estimates
2 samples_sequence,
3 samples_f,
4 ml_subtype,
5 prob_ml_subtype,
6 ml_stage,
7 prob_ml_stage,
----> 8 prob_subtype_stage = sustain.run_sustain_algorithm(plot=True)

File ~/.conda/envs/sustain_env/lib/python3.8/site-packages/pySuStaIn/AbstractSustain.py:186, in AbstractSustain.run_sustain_algorithm(self, plot, plot_format, **kwargs)
178 # max like subtype and stage / subject
179 N_samples = 1000
180 ml_subtype,
181 prob_ml_subtype,
182 ml_stage,
183 prob_ml_stage,
184 prob_subtype,
185 prob_stage,
--> 186 prob_subtype_stage = self.subtype_and_stage_individuals(self.__sustainData, samples_sequence, samples_f, N_samples) #self.subtype_and_stage_individuals(self.__data, samples_sequence, samples_f, N_samples)
187 if not pickle_filepath.exists():
189 if not os.path.exists(self.output_folder):

File ~/.conda/envs/sustain_env/lib/python3.8/site-packages/pySuStaIn/AbstractSustain.py:590, in AbstractSustain.subtype_and_stage_individuals(self, sustainData, samples_sequence, samples_f, N_samples)
587 except:
588 prob_ml_subtype[i] = this_prob_subtype[this_subtype[0][0]]
--> 590 this_prob_stage = np.squeeze(prob_subtype_stage[i, :, int(ml_subtype[i])])
592 if (np.sum(np.isnan(this_prob_stage)) == 0):
593 # this_stage =
594 this_stage = np.where(this_prob_stage == np.max(this_prob_stage))

ValueError: cannot convert float NaN to integer

``

[CNS521666]

还是一个C++语言的问题，不能跑混合效应模型，我也没搞清楚。

[noxtoby]

It's difficult to help you to get to the source of your problem, but it looks like ml_subtype is NaN.

Possibly due to divide-by-zero error somewhere. See also issue 28