Indexing Issue: Multiple Maximum Values in Posterior Distribution Leads to Unexpected Array Size of Index

[kompeki0]

Bug: An error occurs when fitting a specific array under certain conditions.

Environment:

• Python version: Python 3.7.3
• psignifit version: 0.1
• OS: macOS 14.6

Description:
An error occurs when running the following program:

import psignifit as ps
import numpy as np

arr = np.asarray([[0, 1, 10], [4, 0, 10], [8, 6, 10], [12, 8, 10], [16, 10, 10]])
options = {}
options['fixedPars'] = np.ones(5)*np.nan
options['fixedPars'][2] = 0.0
options['fixedPars'][3] = 0.0
options['sigmoidName'] = 'logistic'

result = ps.psignifit(arr, options)

Error Code:

File "psignifit/psignifit.py", line 393, in psignifitCore
    Fit[idx] = result['X1D'][idx][index[idx]] 
ValueError: could not broadcast input array from shape (2,) into shape (1,)

Remarks:
It seems that the issue arises in the following part of psignifit.py:

index = np.where(result['Posterior'] == np.max(result['Posterior'].ravel()))

The problem occurs when result['Posterior'].ravel() contains multiple maximum values, causing index to contain unexpected values (in this case, two values).

To work around this issue, I handled the case where index has multiple values by averaging them if they are consecutive.

Proposed Fix:
I modified the following part:

Fit[idx] = result['X1D'][idx][index[idx]]

to:

if len(index[idx]) > 1:
    unique_sorted_arr = sorted(set(index[idx]))
    expected_sequence = list(range(min(unique_sorted_arr), max(unique_sorted_arr) + 1))
    if unique_sorted_arr == expected_sequence:
        warnings.warn(f"index has multiple consecutive values: {index[idx]}. The mean value of results is used instead", UserWarning)
        Fit[idx] = result['X1D'][idx][index[idx]].mean() 
    else:
        raise ValueError(f"index has multiple non-consecutive values: {index[idx]}.")
else:
    Fit[idx] = result['X1D'][idx][index[idx]]

However, I am not fully confident that this fix is universally applicable.