update to fmri_data.mean to handle selective averaging

CanlabCore/@image_vector/mean.m

@@ -15,6 +15,9 @@
 % - Treats values of 0 as missing (after SPM) and excludes from mean.
 %
 % :Optional Inputs:
+%   - 'group', 'group_by' -> followed by vector of integers to define
+%       groups (e.g., images from the same participant). Averages across images
+%       within each group.
 %   - 'write', followed by file name
 %   - 'path', followed by location for file (default = current directory)
 %   - 'orthviews' -> show orthviews for this image, same as orthviews(m)
@@ -38,6 +41,7 @@
 doplot = false;
 dogroup = false;
 group_by = [];
+u = [];
 
 for i = 1:length(varargin)
     if ischar(varargin{i})
@@ -50,7 +54,7 @@
             case {'hist', 'histogram'}, dohist = 1;
             case {'orth', 'orthviews'}, doorth = 1;
 
-            case 'group'
+            case {'group', 'group_by'}
                 dogroup = true;
                 group_by = varargin{i+1};
                 group_by = categorical(group_by);  % this should work for strings or numeric arrays
@@ -90,7 +94,8 @@
     if ~isempty(obj.metadata_table)
 
         m.metadata_table = obj.metadata_table;
-        m = average_metadata_table(m, group_by, u);
+
+        m = average_metadata_table(m, group_by, u); % can handle single-group case (empty grouping var) or separate groups. Need to do this even if group_by = [] so we end up with 1 row
 
     end
 
@@ -100,16 +105,44 @@
     m = image_vector('dat', mean(obj.dat', 1, 'omitnan')', 'volInfo', obj.volInfo, 'noverbose');
 end
 
+% Copy selected other fields
+% -------------------------------
+copyfield = {'source_notes','images_per_session','Y_names','Y_descrip',...
+    'covariate_names','additional_info','history', 'removed_voxels'};
+for i = 1:length(copyfield)
+    m.(copyfield{i}) = obj.(copyfield{i});
+end
+
+if ~iscell(m.history) || isempty(m.history), m.history = {m.history}; end
+m.history{end + 1} = 'Averaged images';
+if dogroup, m.history{end} = 'Averaged images by group'; end
+
+% Selective average by group of other fields
+% -------------------------------
+u = [];
+m.Y = average_var_by_group(m.Y, group_by, u);                       % can handle single-group case (empty grouping var) or separate groups
+m.covariates = average_var_by_group(m.covariates, group_by, u);     % can handle single-group case (empty grouping var) or separate groups
+m.X = average_var_by_group(m.X, group_by, u);     % can handle single-group case (empty grouping var) or separate groups
+
+% Recast if needed
+% -------------------------------
+if isa(obj, 'fmri_data_st')
+    m = fmri_data_st(m);
+end
+
 % Not completed for statistic_image
 % if isa(obj, 'statistic_image')
 %     m = statistic_image(m);
 % end
 
-m.removed_voxels = obj.removed_voxels;
-
+% Other items
+% -------------------------------
 % convert NaNs in means back to 0s for compatibility
 m.dat(isnan(m.dat)) = 0;
 
+% Plots and other output
+% -------------------------------
+
 if doplot || doorth
     orthviews(m);
 end
@@ -152,9 +185,11 @@
 
 
 function obj = average_metadata_table(obj, group_by, u)
-% average values in table
+
+% average values in entire table if group_by is missing
 if isempty(group_by)
-    group_by = categorical(ones(size(obj.dat, 2), 1));
+    group_by = ones(size(obj.metadata_table, 1), 1);
+    u = 1;
 end
 
 newtable = obj.metadata_table(1, :); % placeholder
@@ -183,3 +218,47 @@
 obj.metadata_table = newtable;
 
 end % function
+
+% NOTES: from fmri_data_st, in case we want to use later for char and to
+% print warnings
+% if ~isempty(t)
+%         tnames = t.Properties.VariableNames;
+%         vars = cell(1,length(tnames));
+%         for i = 1:length(tnames)
+%             if isnumeric(t.(tnames{i}))
+%                 vars{i} = nanmean(t.(tnames{i}));
+%             elseif iscell(t.(tnames{i}))
+%                 vars{i} = t.(tnames{i})(1);
+%             elseif ischar(t.(tnames{i}))
+%                 vars{i} = t.(tnames{i})(1,:);
+%             else
+%                 warning(['No policy for averaging datatype ''' class(t.(tnames{i})) ''' in metadata_table. Dropping column ''' tnames{i} '''']);
+%                 vars{i} = [];
+%             end
+%         end
+
+
+function newv = average_var_by_group(v, group_by, u)  % can handle single-group case (empty grouping var) or separate groups
+% average values in table
+if isempty(group_by)
+    group_by = ones(size(v, 1), 1);
+    u = 1;
+end
+
+newv = NaN * zeros(length(u), size(v, 2));
+
+if isempty(v), return, end
+validateattributes(v, {'numeric'})
+if size(v, 1) ~= length(group_by), error('Variable and group_by vector must be the same length. Check input object fields and grouping variable'); end
+
+
+for i = 1:length(u)
+    % for each group
+
+    wh = group_by == u(i); % this group/participant
+
+    newv(i, :) = nanmean(v(wh, :), 1); 
+
+end
+
+end % function

CanlabCore/Statistics_tools/contrast_code.m

@@ -4,7 +4,8 @@
 %
 %     [vec,outnames] = contrast_code(vec)
 %
-% Changes values to 1, -1, or 0 for contrast coding
+% Attempts to change values to 1, -1, or 0 for contrast coding
+% Success depends on inputs. Check output
 
 outnames = [];
 
@@ -17,6 +18,13 @@
     end
 
 elseif ~isempty(vec)
+
+    if all(vec == 0 | vec == 1) % dummy codes
+
+        vec(vec == 0) = -1;
+
+    else
+
     wh = find(vec > 0);
 
     vec(wh) = 1;
@@ -25,6 +33,8 @@
 
     vec(wh) = -1;
 
+    end
+
 end
 
 return