Added tests for automatic test type inference function

Author: fabriciocravo

.DS_Store

@@ -0,0 +1,33 @@
+function check_multivariate_group(RP)
+%% check_multivariate_group
+% Validates the multivariate grouping against the selected methods.
+%   - If a multivariate method is selected, a grouping must exist.
+%   - When a grouping exists, it must be n_groups x n_var so each column
+%     corresponds to one variable (edge/voxel).
+%
+% **Author**: Fabricio Cravo
+
+    has_group = ~isequaln(RP.multivariate_group, NaN);
+    
+    % A multivariate method needs a grouping to run
+    if RP.has_mvm && ~has_group
+        error('check_multivariate_group:missingGroup', ...
+            ['A multivariate method was selected but no multivariate ' ...
+            'grouping is available. Set RP.atlas_file or provide ' ...
+            'RP.multivariate_group.']);
+    end
+    
+    % No grouping and no multivariate method -> nothing to check
+    if ~has_group
+        return
+    end
+    
+    % Grouping present -> column count must match n_var
+    n_cols = size(RP.multivariate_group, 2);
+    if n_cols ~= RP.n_var
+        error('check_multivariate_group:sizeMismatch', ...
+            ['Multivariate group has %d columns but RP.n_var = %d. ' ...
+            'Each column must correspond to one variable (edge/voxel).'], ...
+            n_cols, RP.n_var);
+    end
+end

multivariate_groupings/check_multivariate_group.m

@@ -0,0 +1,5 @@
+function mvg = mvg_shen_network(Params)
+   
+    mvg = 'Pass';
+
+end

multivariate_groupings/mvg_shen_network.m

@@ -0,0 +1,24 @@
+function [mvg_group, n_groups_mvm] = set_multivariate_group(RP)
+
+    % Explicit matrix supplied -> use as-is
+    if ~isequaln(RP.multivariate_group, NaN)
+        mvg_group = RP.multivariate_group;
+        n_groups_mvm = size(mvg_group, 1);
+        return
+    end
+
+    % No atlas / no groups -> nothing to build
+    if isempty(RP.edge_groups) || RP.n_networks == 0
+        mvg_group = NaN;
+        n_groups_mvm = 0;
+        return
+    end
+
+    % Reuse the same flatten that builds Y's rows -> guaranteed alignment
+    edge_labels = RP.flat_matrix_fun(RP.edge_groups);   % same order as Y rows
+    edge_labels = edge_labels(:)';
+    group_ids   = unique(edge_labels(edge_labels > 0)); % drop 0 = unassigned
+    mvg_group   = sparse(group_ids(:) == edge_labels);        % n_groups x n_edges logical
+    n_groups_mvm = numel(group_ids);
+
+end

multivariate_groupings/set_multivariate_group.m

@@ -35,11 +35,12 @@
 % tests subs_data_from_score_condition function
 test_subs_score_retrieval()
 
+% Test automatic test type retrieval function
+test_type_inference_tests()
+
 % test if calculation status is proper
 test_check_calculation_status()
 
-
-
 clean_test_directories()
 
 create_test_fc_data_set()

power_calculator_test_script.m

@@ -1,4 +1,4 @@
-function Params = check_stat_method_class_validity(Params)
+function [Params, has_mvm] = check_stat_method_class_validity(Params)
 % check_stat_method_class_validity - Validates properties of statistical method classes.
 %
 % Checks that method classes in Params.all_cluster_stat_types:
@@ -96,7 +96,18 @@
             error('Method %s has more local permutations defined than can be prepared by setparams', method_name)
         end
     end
+    
+    %% Check if multivariate method exists
+    
+    has_mvm = false;
+    for i = 1:numel(Params.all_cluster_stat_types)
+        method_obj = feval(Params.all_cluster_stat_types{i});
 
+        if strcmp(method_obj.level,'multivariate')
+            has_mvm = true;
+            break
+        end
 
+    end
 
 end

power_calculator_tools/check_stat_method_class_validity.m

@@ -10,6 +10,9 @@
 
         case 'network'
             level = 'network';
+
+        case 'multivariate'
+            level = 'multivariate';
 
         case 'node'
             level = 'variable';

power_calculator_tools/extract_stat_level.m

@@ -42,9 +42,10 @@
 
     test_score_set = get_test_score_set(TestData);
 
-    if length(test_score_set) == 1 && ~isnan(test_score_set)
+    if length(test_score_set) == 1 
         % if all scores are equal to the same number - t test
         test_type = 't';
+        test_type_origin = 'score_cond';
 
     elseif length(test_score_set) > 2
         % if score is continuous -> r
@@ -59,9 +60,18 @@
         [index_cond_1, index_cond_2] = get_index_matching_score(TestData.score, ...
             test_score_set);
 
-        sub_ids_cond1 = BrainData.(TestData.reference_condition).sub_ids(index_cond_1);
-        sub_ids_cond2 = BrainData.(TestData.reference_condition).sub_ids(index_cond_2);
+        % resolve to actual subject IDs in TestData's own frame first
+        sub_ids_test_cond1 = TestData.sub_ids(index_cond_1);
+        sub_ids_test_cond2 = TestData.sub_ids(index_cond_2);
+
+        % now cross into BrainData's via ID matching
+        index_b_data_c1 = ismember(BrainData.(TestData.reference_condition).sub_ids, sub_ids_test_cond1);
+        index_b_data_c2 = ismember(BrainData.(TestData.reference_condition).sub_ids, sub_ids_test_cond2);
 
+        % Get the ids for the condition
+        sub_ids_cond1 = BrainData.(TestData.reference_condition).sub_ids(index_b_data_c1);
+        sub_ids_cond2 = BrainData.(TestData.reference_condition).sub_ids(index_b_data_c2);
+
         %% TODO: Divided by the two - focus on group sizes
         n_equal = numel(intersect(sort(sub_ids_cond1), sort(sub_ids_cond2)));
         n_unique = numel(setxor(sub_ids_cond1, sub_ids_cond2));

power_calculator_tools/infer_test_from_data.m

@@ -59,6 +59,8 @@
                         method_struct.(full_method_name) = zeros(length(unique(RP.edge_groups)) - 1, RP.n_repetitions);
                     case "variable"
                         method_struct.(full_method_name) = zeros(RP.n_var, RP.n_repetitions);
+                    case "multivariate"
+                        method_struct.(full_method_name) = zeros(RP.n_multivariate, RP.n_repetitions);
                     otherwise
                         error("Unknown statistic level: %s", method_instance.level);
                 end

power_calculator_tools/initialize_global_pvals.m

@@ -92,6 +92,7 @@ function process_repetition_batches(X, Y, RP, UI)
         STATS.is_permutation_based = RP.is_permutation_based; 
         STATS.thresh = RP.tthresh_first_level;
         STATS.alpha = RP.pthresh_second_level;
+        STATS.multivariate_group = RP.multivariate_group;
 
         % **Loop through missing repetitions**
         STATSc = parallel.pool.Constant(STATS);