Added dataset assertion to check data integrety before running calcs'

Author: fabriciocravo

.DS_Store

@@ -0,0 +1,50 @@
+function assert_dataset(Dataset)
+    % assert_dataset
+    % Validates a PRISME dataset before it enters the pipeline. Fails loud at
+    % ingestion rather than letting bad values propagate into curve fits.
+    %   1. brain_data has no NaN (and no Inf)
+    %   2. brain_data.data is a proper 2-D edge × subject matrix
+    %   3. sub_ids length matches the number of data columns (subjects)
+    
+    assert(isfield(Dataset, 'brain_data'), ...
+        'assert_dataset: missing brain_data field');
+    
+    conds = fieldnames(Dataset.brain_data);
+    assert(~isempty(conds), 'assert_dataset: brain_data has no conditions');
+    
+    for c = 1:numel(conds)
+        cond = conds{c};
+        S = Dataset.brain_data.(cond);
+    
+        assert( ...
+            isfield(S, 'data') && isfield(S, 'sub_ids'), ...
+            'assert_dataset: %s missing data or sub_ids field', ...
+            cond ...
+        );
+    
+        d = S.data;
+    
+        [n_vars, n_subj] = size(d);
+        n_ids = numel(S.sub_ids);
+
+        % Subjects are the COLUMNS, defined by sub_ids. Check that the column
+        % count matches, and that edges (rows) outnumber subjects as expected.
+        assert(n_subj == n_ids, ...
+            ['assert_dataset: %s has %d sub_ids but %d data columns — ' ...
+            'subject labels and brain data are misaligned (or matrix transposed: ' ...
+            'data is %d x %d)'], cond, n_ids, n_subj, n_vars, n_subj);
+    
+        % --- Check 1: no NaN, no Inf ---
+        n_nan = sum(isnan(d(:)));
+        n_inf = sum(isinf(d(:)));
+        if n_nan > 0 || n_inf > 0
+            per_subject = sum(isnan(d), 1);
+            bad = find(per_subject > 0);
+            error('assert_dataset:badValues', ...
+                '%s.data has %d NaN and %d Inf across %d subjects: [%s]', ...
+                cond, n_nan, n_inf, numel(bad), num2str(bad));
+        end
+
+    end
+
+end

assertion_scripts/assert_dataset.m

@@ -0,0 +1,47 @@
+% clean_rosenblatt_nans.m
+% Standalone script: loads s_slim_fc_rosenblatt.mat, removes any subject whose
+% brain data contains a NaN (across all conditions under brain_data), removes the
+% matching entries from sub_ids, sub_ids_motion, and motion, and saves the result
+% as s_slim_fc_rosenblatt_fabi.mat in the same directory.
+
+clear; clc;
+
+in_file  = '/Users/f.cravogomes/Desktop/Cloned Repos/PRISME-Brain-Power-Calculator/data/s_slim_fc_rosenblatt.mat';
+out_file = '/Users/f.cravogomes/Desktop/Cloned Repos/PRISME-Brain-Power-Calculator/data/s_slim_fc_rosenblatt_fabi.mat';
+
+fprintf('Loading %s\n', in_file);
+Dataset = load(in_file);
+outcome = Dataset.outcome;
+study_info = Dataset.study_info;
+
+conds = fieldnames(Dataset.brain_data);
+
+for c = 1:numel(conds)
+    cond = conds{c};
+    S = Dataset.brain_data.(cond);
+
+    % data is edge × subject: a bad subject is any column containing a NaN
+    good  = ~any(isnan(S.data), 1);        % 1 × n_subjects logical
+    n_bad = sum(~good);
+
+    if n_bad == 0
+        fprintf('%s: no NaN subjects, unchanged (%d subjects)\n', cond, numel(good));
+        continue;
+    end
+
+    % Subset every subject-aligned field by the SAME column selection
+    S.data           = S.data(:, good);
+    S.sub_ids        = S.sub_ids(good);
+    S.sub_ids_motion = S.sub_ids_motion(good);
+    S.motion         = S.motion(good);
+
+    assert(~any(isnan(S.data(:))), '%s: NaN remaining after removal', cond);
+
+    brain_data.(cond) = S;
+end
+
+% Save under the ORIGINAL variable name so downstream loaders still work.
+fprintf('Saving %s\n', out_file);
+save(out_file, 'brain_data', 'outcome', 'study_info', '-v7.3');
+
+fprintf('Done.\n');

atlas_storage/test_hcp_fc_atlas.mat

@@ -0,0 +1,14 @@
+function transpose_ukb_brain_data()
+    origin = '/Users/f.cravogomes/Desktop/Cloned Repos/PRISME-Brain-Power-Calculator/data/s_ukb_fc_jiang.mat';
+    dest   = '/Users/f.cravogomes/Desktop/Cloned Repos/PRISME-Brain-Power-Calculator/data/s_ukb_fc_fabricio.mat';
+    
+    S = load(origin);
+    
+    conds = fieldnames(S.brain_data);
+    for c = 1:numel(conds)
+        S.brain_data.(conds{c}).data = S.brain_data.(conds{c}).data.';
+    end
+    
+    builtin('save', dest, '-struct', 'S', '-v7.3');
+
+end