[BUG] cannot use dataframe as input for second level GLM #3871

gurraburra · 2023-08-01T12:47:39Z

Is there an existing issue for this?

I have searched the existing issues

Operating system

Linux
Mac
Windows

Operating system version

Mac OS Version 13.4.1 "ventura"

Python version

nilearn version

0.10.1

Expected behavior

Using unlearn.glm.second_level.SecondLevelModel with panda.DataFrame as input I get an error when computing the contrast because the code checks the first element of the second_level_input to see if it is a FirstLevelModel. The problem is when the second_level_input is a DataFrame, a key error is generated. See below:

Current behavior & error messages

This is what I got:

KeyError                                  Traceback (most recent call last)
File [/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/pandas/core/indexes/base.py:3652](https://file+.vscode-resource.vscode-cdn.net/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/pandas/core/indexes/base.py:3652), in Index.get_loc(self, key)
   3651 try:
-> 3652     return self._engine.get_loc(casted_key)
   3653 except KeyError as err:

File [/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/pandas/_libs/index.pyx:147](https://file+.vscode-resource.vscode-cdn.net/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/pandas/_libs/index.pyx:147), in pandas._libs.index.IndexEngine.get_loc()

File [/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/pandas/_libs/index.pyx:176](https://file+.vscode-resource.vscode-cdn.net/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/pandas/_libs/index.pyx:176), in pandas._libs.index.IndexEngine.get_loc()

File pandas/_libs/hashtable_class_helper.pxi:7080, in pandas._libs.hashtable.PyObjectHashTable.get_item()

File pandas/_libs/hashtable_class_helper.pxi:7088, in pandas._libs.hashtable.PyObjectHashTable.get_item()

KeyError: 0

The above exception was the direct cause of the following exception:

KeyError                                  Traceback (most recent call last)
Cell In[169], line 1
----> 1 slm.compute_contrast(first_level_contrast="insula-l")

File [/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/nilearn/glm/second_level/second_level.py:543](https://file+.vscode-resource.vscode-cdn.net/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/nilearn/glm/second_level/second_level.py:543), in SecondLevelModel.compute_contrast(self, second_level_contrast, first_level_contrast, second_level_stat_type, output_type)
    540     raise ValueError("The model has not been fit yet")
    542 # check first_level_contrast
--> 543 _check_first_level_contrast(
    544     self.second_level_input_, first_level_contrast
    545 )
    547 # check contrast and obtain con_val
    548 con_val = _get_con_val(second_level_contrast, self.design_matrix_)

File [/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/nilearn/glm/second_level/second_level.py:181](https://file+.vscode-resource.vscode-cdn.net/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/nilearn/glm/second_level/second_level.py:181), in _check_first_level_contrast(second_level_input, first_level_contrast)
    180 def _check_first_level_contrast(second_level_input, first_level_contrast):
--> 181     if isinstance(second_level_input[0], FirstLevelModel):
    182         if first_level_contrast is None:
    183             raise ValueError(
    184                 "If second_level_input was a list of "
    185                 "FirstLevelModel, then first_level_contrast "
   (...)
    188                 "compute_contrast method of FirstLevelModel"
    189             )

File [/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/pandas/core/frame.py:3761](https://file+.vscode-resource.vscode-cdn.net/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/pandas/core/frame.py:3761), in DataFrame.__getitem__(self, key)
   3759 if self.columns.nlevels > 1:
   3760     return self._getitem_multilevel(key)
-> 3761 indexer = self.columns.get_loc(key)
   3762 if is_integer(indexer):
   3763     indexer = [indexer]

File [/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/pandas/core/indexes/base.py:3654](https://file+.vscode-resource.vscode-cdn.net/opt/homebrew/Caskroom/miniforge/base/envs/rs-analysis/lib/python3.11/site-packages/pandas/core/indexes/base.py:3654), in Index.get_loc(self, key)
   3652     return self._engine.get_loc(casted_key)
   3653 except KeyError as err:
-> 3654     raise KeyError(key) from err
   3655 except TypeError:
   3656     # If we have a listlike key, _check_indexing_error will raise
   3657     #  InvalidIndexError. Otherwise we fall through and re-raise
   3658     #  the TypeError.
   3659     self._check_indexing_error(key)

KeyError: 0

Steps and code to reproduce bug

# Paste your code here
import pandas as pd
from nilearn.glm.second_level import SecondLevelModel

# need a random file tmp.nii.gz
df = pd.DataFrame(data={"effects_map_path" : ["tmp.nii.gz"], "map_name" : ["tmp"], "subject_label" : ["sub1"]})
slm = SecondLevelModel().fit(second_level_input=df)
slm.compute_contrast()

Remi-Gau · 2023-08-01T21:18:57Z

I think this reproduces the bug

import pandas as pd
from nilearn.glm.second_level import SecondLevelModel
from nilearn._utils.data_gen import write_fake_fmri_data_and_design

# need a random file fmri_run0.nii
write_fake_fmri_data_and_design(shapes=((7, 8, 7, 15), (7, 8, 7, 16)))
df = pd.DataFrame(data={"effects_map_path" : ["fmri_run0.nii"], "map_name" : ["tmp"], "subject_label" : ["sub1"]})
slm = SecondLevelModel().fit(second_level_input=df)
slm.compute_contrast()

Traceback (most recent call last):
  File "/home/remi/github/nilearn/env/lib/python3.11/site-packages/pandas/core/indexes/base.py", line 3653, in get_loc
    return self._engine.get_loc(casted_key)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "pandas/_libs/index.pyx", line 147, in pandas._libs.index.IndexEngine.get_loc
  File "pandas/_libs/index.pyx", line 176, in pandas._libs.index.IndexEngine.get_loc
  File "pandas/_libs/hashtable_class_helper.pxi", line 7080, in pandas._libs.hashtable.PyObjectHashTable.get_item
  File "pandas/_libs/hashtable_class_helper.pxi", line 7088, in pandas._libs.hashtable.PyObjectHashTable.get_item
KeyError: 0

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/home/remi/github/nilearn/tmp.py", line 10, in <module>
    slm.compute_contrast()
  File "/home/remi/github/nilearn/nilearn/glm/second_level/second_level.py", line 544, in compute_contrast
    _check_first_level_contrast(
  File "/home/remi/github/nilearn/nilearn/glm/second_level/second_level.py", line 182, in _check_first_level_contrast
    if isinstance(second_level_input[0], FirstLevelModel):
                  ~~~~~~~~~~~~~~~~~~^^^
  File "/home/remi/github/nilearn/env/lib/python3.11/site-packages/pandas/core/frame.py", line 3761, in __getitem__
    indexer = self.columns.get_loc(key)
              ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/remi/github/nilearn/env/lib/python3.11/site-packages/pandas/core/indexes/base.py", line 3655, in get_loc
    raise KeyError(key) from err
KeyError: 0

Remi-Gau · 2023-08-01T21:37:30Z

Went and looked into our test suite to see how we test this and here is something that works

import pandas as pd
from nilearn.glm.second_level import SecondLevelModel
from nilearn._utils.data_gen import write_fake_fmri_data_and_design

shapes = ((7, 8, 9, 10),)
_, FUNCFILE, _ = write_fake_fmri_data_and_design(shapes)
FUNCFILE = FUNCFILE[0]

dfcols = ["subject_label", "map_name", "effects_map_path"]
dfrows = [
    ["01", "a", FUNCFILE],
    ["02", "a", FUNCFILE],
    ["03", "a", FUNCFILE],
]
niidf = pd.DataFrame(dfrows, columns=dfcols)

# dataframes as input
SecondLevelModel().fit(niidf)

Remi-Gau · 2023-08-01T21:39:13Z

extracted from here:

nilearn/nilearn/glm/tests/test_second_level.py

Line 511 in a95149b

def test_fmri_inputs():

Remi-Gau · 2023-08-01T21:50:16Z

Went and looked into our test suite to see how we test this and here is something that works

import pandas as pd
from nilearn.glm.second_level import SecondLevelModel
from nilearn._utils.data_gen import write_fake_fmri_data_and_design

shapes = ((7, 8, 9, 10),)
_, FUNCFILE, _ = write_fake_fmri_data_and_design(shapes)
FUNCFILE = FUNCFILE[0]

dfcols = ["subject_label", "map_name", "effects_map_path"]
dfrows = [
    ["01", "a", FUNCFILE],
    ["02", "a", FUNCFILE],
    ["03", "a", FUNCFILE],
]
niidf = pd.DataFrame(dfrows, columns=dfcols)

# dataframes as input
SecondLevelModel().fit(niidf)

my bad this does not work when you compute a contrast on the model

bthirion · 2023-08-02T21:15:39Z

Thx for reporting.
Do you know how to fix that ?

gurraburra · 2023-08-02T21:25:21Z

It seems that the bug comes from the call to _check_first_level_contrast, the function tries to check if the first element of second_level_input is an instance of FirstLevelModel since than the first_level_contrast argument is needed to compute the contrast. An easy fix would just to first check if the second_level_input is a list:
if isinstance(second_level_input, list) and isinstance(second_level_input[0], FirstLevelModel):

bthirion · 2023-08-03T07:33:17Z

OK. Do you want to contribute it ?

gurraburra added the Bug for bug reports label Aug 1, 2023

Remi-Gau changed the title ~~[BUG]~~ [BUG] cannot use dataframe as input for second level GLM Aug 1, 2023

Remi-Gau mentioned this issue Aug 3, 2023

[FIX] ensure that dataframes can be used as input for second level GLM (estimation and contrast) #3879

Merged

Remi-Gau closed this as completed in #3879 Aug 7, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[BUG] cannot use dataframe as input for second level GLM #3871

[BUG] cannot use dataframe as input for second level GLM #3871

gurraburra commented Aug 1, 2023

Remi-Gau commented Aug 1, 2023

Remi-Gau commented Aug 1, 2023

Remi-Gau commented Aug 1, 2023

Remi-Gau commented Aug 1, 2023

bthirion commented Aug 2, 2023

gurraburra commented Aug 2, 2023

bthirion commented Aug 3, 2023

[BUG] cannot use dataframe as input for second level GLM #3871

[BUG] cannot use dataframe as input for second level GLM #3871

Comments

gurraburra commented Aug 1, 2023

Is there an existing issue for this?

Operating system

Operating system version

Python version

nilearn version

Expected behavior

Current behavior & error messages

Steps and code to reproduce bug

Remi-Gau commented Aug 1, 2023

Remi-Gau commented Aug 1, 2023

Remi-Gau commented Aug 1, 2023

Remi-Gau commented Aug 1, 2023

bthirion commented Aug 2, 2023

gurraburra commented Aug 2, 2023

bthirion commented Aug 3, 2023