Merge pull request #387 from effigies/dx/pybids

FIX: Handle correct contrasts and intercepts being passed from PyBIDS

fitlins/cli/run.py

@@ -261,6 +261,7 @@ def get_parser():
 
 def run_fitlins(argv=None):
     import re
+    import nipype
     from nipype import logging as nlogging
 
     warnings.showwarning = _warn_redirect
@@ -279,6 +280,9 @@ def run_fitlins(argv=None):
         for ign in opts.ignore or ()
     ]
 
+    if opts.debug:
+        nipype.config.set('execution', 'remove_unnecessary_outputs', False)
+
     log_level = 25 + 5 * (opts.quiet - opts.verbose)
     logger.setLevel(log_level)
     nlogging.getLogger('nipype.workflow').setLevel(log_level)

fitlins/conftest.py

@@ -42,7 +42,7 @@ def sample_model_dict():
                 },
                 "Model": {"X": ["trial_type.ice_cream", "trial_type.cake", "food_sweats"]},
                 "DummyContrasts": {
-                    "Conditions": ["trial_type.ice_cream", "trial_type.cake"],
+                    "Contrasts": ["trial_type.ice_cream", "trial_type.cake"],
                     "Test": "t",
                 },
                 "Contrasts": [
@@ -69,13 +69,16 @@ def sample_model_dict():
             },
             {
                 "Level": "dataset",
-                "Name": "all_food_good_food",
+                "Name": "dataset",
                 "GroupBy": ["contrast"],
-                "Model": {"X": [1]},
-                "DummyContrasts": {
-                    "Conditions": ["icecream_gt_cake", "eating_vs_baseline"],
-                    "Test": "t",
-                },
+                "Model": {"X": [1], "Type": "glm"},
+                "DummyContrasts": {"Test": "t"},
+            },
+            {
+                "Level": "dataset",
+                "Name": "all_food_good_food",
+                "GroupBy": [],
+                "Model": {"X": ["trial_type.ice_cream", "trial_type.cake"], "Type": "glm"},
                 "Contrasts": [
                     {
                         "Name": "all_food_good_food",
@@ -90,9 +93,14 @@ def sample_model_dict():
             {"Source": "run", "Destination": "subject"},
             {
                 "Source": "subject",
-                "Destination": "all_food_good_food",
+                "Destination": "dataset",
                 "Filter": {"contrast": ["icecream_gt_cake", "eating_vs_baseline"]},
             },
+            {
+                "Source": "subject",
+                "Destination": "all_food_good_food",
+                "Filter": {"contrast": ["trial_type.ice_cream", "trial_type.cake"]},
+            },
         ],
     }
 

fitlins/interfaces/afni.py

@@ -338,6 +338,8 @@ def save_tsnr(self, runtime, rbetas, rvars):
         # find the name of the constant column
         if 'constant' in mat.columns:
             const_name = 'constant'
+        elif 'intercept' in mat.columns:
+            const_name = 'intercept'
         else:
             const_name = mat.columns[np.isclose(mat, 1).all(0)].values[0]
         const_idx = np.where(np.array(vol_labels) == const_name)[0]

fitlins/interfaces/nistats.py

@@ -128,6 +128,13 @@ def _run_interface(self, runtime):
             add_reg_names=column_names,
             drift_model=drift_model,
         )
+        # Can get two intercepts from input and nilearn ("constant")
+        # Normalize to "intercept" unless that exists and is really non-1
+        # Force to be exactly 1, in case of weird rounding errors
+        intercept = 'intercept' in mat or 'constant' in mat
+        mat.drop(columns=['intercept', 'constant'], errors='ignore', inplace=True)
+        if intercept:
+            mat['intercept'] = np.ones(mat.shape[0], dtype='u1')
 
         mat.to_csv('design.tsv', sep='\t')
         self._results['design_matrix'] = os.path.join(runtime.cwd, 'design.tsv')

fitlins/interfaces/visualizations.py

@@ -115,7 +115,7 @@ def _visualize(self, data, out_name):
         plot_and_save(
             out_name,
             plot_corr_matrix,
-            data.drop(columns='constant', errors='ignore').corr(),
+            data.drop(columns=['intercept', 'constant'], errors='ignore').corr(),
             len(evs),
         )
 

setup.cfg

@@ -29,7 +29,7 @@ install_requires =
     nilearn~=0.9.1
     pandas>=0.19
     tables>=3.2.1
-    pybids~=0.15.1
+    pybids~=0.15.4
     jinja2
 
 [options.extras_require]