refactor

.github/workflows/run_tests_cli.yml

@@ -60,10 +60,10 @@ jobs:
                 coverage run --source src -m pytest
                 coverage xml
 
-        -   name: Code coverage
-            uses: codecov/codecov-action@v4
-            with:
-                file: coverage.xml
-                flags: cli
-                name: codecov-cli
-                fail_ci_if_error: false
+        # -   name: Code coverage
+        #     uses: codecov/codecov-action@v4
+        #     with:
+        #         file: coverage.xml
+        #         flags: cli
+        #         name: codecov-cli
+        #         fail_ci_if_error: false

miss_hit.cfg

@@ -37,6 +37,6 @@ tab_width: 2
 
 # metrics limit for the code quality (https://florianschanda.github.io/miss_hit/metrics.html)
 metric "cnest": limit 5
-metric "file_length": limit 800
+metric "file_length": limit 1000
 metric "cyc": limit 22
 metric "parameters": limit 7

src/batches/stats/setBatchFactorialDesign.m

@@ -64,17 +64,11 @@
       % TODO all contrasts should have the same name
       contrasts = getContrastsListForFactorialDesign(opt, nodeName);
 
-      % TODO refactor
-      designMatrix = opt.model.bm.get_design_matrix('Name', nodeName);
-      designMatrix = cellfun(@(x) num2str(x), designMatrix, 'uniformoutput', false);
-
-      groupColumnHdr = setxor(designMatrix, {'1'});
-      groupColumnHdr = groupColumnHdr{1};
-
       % Sorting is important so that we know in which order
       % the groups are entered in the design matrix.
       % Otherwise it will be harder to properly design
       % the contrast vectors later.
+      groupColumnHdr = opt.model.bm.getGroupColumnHdrFromDesignMatrix(nodeName);
       availableGroups = getAvailableGroups(opt, groupColumnHdr);
 
       label = '1WayANOVA';
@@ -101,10 +95,9 @@
         % Note that this will lead to different results
         % depending on the requested subejcts
         %
-        tsvFile = fullfile(opt.dir.raw, 'participants.tsv');
-        tsv = bids.util.tsvread(tsvFile);
-        subjectsInGroup = strcmp(tsv.(groupColumnHdr), thisGroup);
-        subjectsLabel = regexprep(tsv.participant_id(subjectsInGroup), '^sub-', '');
+        participants = bids.util.tsvread(fullfile(opt.dir.raw, 'participants.tsv'));
+        subjectsInGroup = strcmp(participants.(groupColumnHdr), thisGroup);
+        subjectsLabel = regexprep(participants.participant_id(subjectsInGroup), '^sub-', '');
         subjectsLabel = intersect(subjectsLabel, opt.subjects);
 
         % collect all con images from all subjects
@@ -183,13 +176,8 @@
 
   contrasts = getContrastsListForFactorialDesign(opt, nodeName);
 
-  node = opt.model.bm.get_nodes('Name', nodeName);
-  groupBy = node.GroupBy;
-
-  % TODO refactor
-  groupColumnHdr = setxor(groupBy, {'contrast'});
-  groupColumnHdr = groupColumnHdr{1};
-
+  participants = bids.util.tsvread(fullfile(opt.dir.raw, 'participants.tsv'));
+  groupColumnHdr = opt.model.bm.getGroupColumnHdrFromGroupBy(nodeName, participants);
   availableGroups = getAvailableGroups(opt, groupColumnHdr);
 
   for iGroup = 1:numel(availableGroups)
@@ -202,10 +190,8 @@
     % Note that this will lead to different results depending on the requested
     % subejcts
     %
-    tsvFile = fullfile(opt.dir.raw, 'participants.tsv');
-    tsv = bids.util.tsvread(tsvFile);
-    subjectsInGroup = strcmp(tsv.(groupColumnHdr), thisGroup);
-    subjectsLabel = regexprep(tsv.participant_id(subjectsInGroup), '^sub-', '');
+    subjectsInGroup = strcmp(participants.(groupColumnHdr), thisGroup);
+    subjectsLabel = regexprep(participants.participant_id(subjectsInGroup), '^sub-', '');
     subjectsLabel = intersect(subjectsLabel, opt.subjects);
 
     % collect all con images from all subjects
@@ -301,6 +287,7 @@
   % If 2 groups, then number of levels = 2
   factorialDesign.des.fd.fact.name = thisGroup;
   factorialDesign.des.fd.fact.levels = numel(icell);
+
   factorialDesign.des.fd.fact.dept = 0;
 
   % 1: Assumes that the variance is not the same across groups
@@ -338,11 +325,11 @@
 
   factorialDesign.masking.tm.tm_none = 1;
   factorialDesign.masking.im = 1;
+
   factorialDesign.masking.em = {''};
 
-  factorialDesign.globalc.g_omit = 1;
-  factorialDesign.globalm.gmsca.gmsca_no = 1;
-  factorialDesign.globalm.glonorm = 1;
+  factorialDesign = setBatchFactorialDesignImplicitMasking(factorialDesign);
+  factorialDesign = setBatchFactorialDesignGlobalCalcAndNorm(factorialDesign);
 
   matlabbatch{end + 1}.spm.stats.factorial_design = factorialDesign;
 
@@ -354,11 +341,12 @@
 
   % TODO refactor
   participants = bids.util.tsvread(fullfile(opt.dir.raw, 'participants.tsv'));
-  columns = fieldnames(participants);
 
-  status = opt.model.bm.validateGroupBy(nodeName, columns);
+  model = opt.model.bm;
+
+  status = model.validateGroupBy(nodeName, participants);
 
-  [glmType, ~, groupBy] = groupLevelGlmType(opt, nodeName, participants);
+  [glmType, groupBy] = model.groupLevelGlmType(nodeName, participants);
 
   % only certain type of model supported for now
   if ismember(glmType, {'unknown', 'two_sample_t_test'})
@@ -370,8 +358,8 @@
     return
   end
 
-  datasetLvlContrasts = opt.model.bm.get_contrasts('Name', nodeName);
-  datasetLvlDummyContrasts = opt.model.bm.get_dummy_contrasts('Name', nodeName);
+  datasetLvlContrasts = model.get_contrasts('Name', nodeName);
+  datasetLvlDummyContrasts = model.get_dummy_contrasts('Name', nodeName);
 
   if isempty(datasetLvlContrasts) && isempty(datasetLvlDummyContrasts)
     msg = sprintf('No contrast specified %s', commonMsg);

src/batches/stats/setBatchGroupLevelContrasts.m

@@ -26,17 +26,21 @@
 
   printBatchName('group level contrast estimation', opt);
 
-  % TODO refactor
+  model = opt.model.bm;
+
   participants = bids.util.tsvread(fullfile(opt.dir.raw, 'participants.tsv'));
-  [groupGlmType, designMatrix, groupBy] =  groupLevelGlmType(opt, nodeName, participants);
 
+  groupColumnHdr = model.getGroupColumnHdrFromGroupBy(nodeName, participants);
+  availableGroups = getAvailableGroups(opt, groupColumnHdr);
+
+  [groupGlmType, groupBy] =  model.groupLevelGlmType(nodeName, participants);
   switch groupGlmType
 
     case 'one_sample_t_test'
 
       contrastsList = getContrastsListForFactorialDesign(opt, nodeName);
 
-      if all(ismember(lower(groupBy), {'contrast'}))
+      if numel(groupBy) == 1 && ismember(lower(groupBy), {'contrast'})
 
         for j = 1:numel(contrastsList)
 
@@ -48,11 +52,7 @@
 
         end
 
-        % TODO make more general than just with group
-      elseif all(ismember(lower(groupBy), {'contrast', 'group'}))
-        % TODO make more general than just with group
-        groupColumnHdr = groupBy{ismember(lower(groupBy), {'group'})};
-        availableGroups = unique(participants.(groupColumnHdr));
+      elseif numel(groupBy) == 2 && any(ismember(groupBy, fieldnames(participants)))
 
         for j = 1:numel(contrastsList)
 
@@ -80,14 +80,10 @@
       % through the Edge filter.
       % Then generate the between group contrasts.
 
-      designMatrix = removeIntercept(designMatrix);
-
-      groups = unique(participants.(designMatrix{1}));
-
-      edge = opt.model.bm.get_edge('Destination', nodeName);
+      edge = model.get_edge('Destination', nodeName);
       contrastsList = edge.Filter.contrast;
 
-      thisContrast = opt.model.bm.get_contrasts('Name', nodeName);
+      thisContrast = model.get_contrasts('Name', nodeName);
 
       for j = 1:numel(contrastsList)
 
@@ -102,14 +98,14 @@
           % Sort conditions and weights
           [ConditionList, I] = sort(thisContrast{iCon}.ConditionList);
           for iCdt = 1:numel(ConditionList)
-            ConditionList{iCdt} =  strrep(ConditionList{iCdt}, [designMatrix{1}, '.'], '');
+            ConditionList{iCdt} =  strrep(ConditionList{iCdt}, [groupColumnHdr, '.'], '');
           end
           Weights = thisContrast{iCon}.Weights(I);
 
           % Create contrast vectors by what was passed in the model
-          convec = zeros(size(groups));
-          for iGroup = 1:numel(groups)
-            index = strcmp(groups{iGroup}, ConditionList);
+          convec = zeros(size(availableGroups));
+          for iGroup = 1:numel(availableGroups)
+            index = strcmp(availableGroups{iGroup}, ConditionList);
             if any(index)
               convec(iGroup) = Weights(index);
             end
@@ -128,17 +124,16 @@
 
       designMatrix = removeIntercept(designMatrix);
 
-      % TODO make more general than just with group
-      if ismember(lower(designMatrix), {'group'})
+      if any(ismember(designMatrix, fieldnames(participants)))
         % TODO will this ignore the contrasts define at other levels
         % and not passed through the filter ?
-        edge = opt.model.bm.get_edge('Destination', nodeName);
+        edge = model.get_edge('Destination', nodeName);
         contrastsList = edge.Filter.contrast;
       end
 
       for j = 1:numel(contrastsList)
 
-        thisContrast = opt.model.bm.get_contrasts('Name', nodeName);
+        thisContrast = model.get_contrasts('Name', nodeName);
 
         spmMatFile = fullfile(getRFXdir(opt, nodeName, contrastsList{j}), 'SPM.mat');
 

src/batches/stats/setBatchSubjectLevelContrasts.m

@@ -4,7 +4,7 @@
   %
   % USAGE::
   %
-  %   matlabbatch = setBatchSubjectLevelContrasts(matlabbatch, opt, subLabel, funcFWHM)
+  %   matlabbatch = setBatchSubjectLevelContrasts(matlabbatch, opt, subLabel, nodeName)
   %
   % :param matlabbatch:
   % :type matlabbatch: structure

src/batches/stats/setBatchTwoSampleTTest.m

@@ -66,9 +66,7 @@
   group1 = group1{2};
   group2 = group2{2};
 
-  % TODO refactor
   availableGroups = getAvailableGroups(opt, groupField);
-  participants = bids.util.tsvread(fullfile(opt.dir.raw, 'participants.tsv'));
 
   if any(~ismember({group1, group2}, availableGroups))
     error(['Some requested group is not present: %s.', ...
@@ -85,6 +83,7 @@
     conImages{iSub} = findSubjectConImage(opt, subLabel, contrastsList);
   end
 
+  participants = bids.util.tsvread(fullfile(opt.dir.raw, 'participants.tsv'));
   % set up the batch
   for iCon = 1:numel(contrastsList)