Cross-validated RSA and 'other_average'

1

Dear TDT experts,

I am trying to run a cross-validated RSA for a design with 4 conditions and 4 runs (all conditions present on each run) in a leave-one-run-out procedure. Specifying ‘other_average’ as output I was expecting to get a 4x4 matrix (4 conditions averaged across 4 runs) as a result, or even a 16x16 matrix (if no averaging across runs takes place, as explained here) but I keep getting a 12x4 matrix. Given the 12x4 size I’m suspecting this is the resulting matrix from a cross-validation fold instead of the average across folds.
The code I’ve been using (based on the similarity_template) is:

cfg = decoding_defaults;
  
cfg.analysis = 'searchlight';
  
cfg.results.dir = out_dir;
beta_loc = beta_dir;
cfg.files.mask = mask_dir;
  
labelnames = {'c1_s1',...
              'c1_s2',...
              'c2_s1',...
              'c2_s2',...
              };

labels = [1 2 3 4];

cfg.decoding.software = 'similarity';
cfg.decoding.method = 'classification';
cfg.decoding.train.classification.model_parameters = 'pearson';
  
cfg.results.output = {'other_average'};
  
cfg.searchlight.unit = 'mm';
cfg.searchlight.radius = searchlight_rad;
cfg.searchlight.spherical = 0;
cfg.verbose = 2; 
cfg.scale.method = 'min0max1';
cfg.scale.estimation = 'all';
cfg.plot_selected_voxels = 100;
  
regressor_names = design_from_spm(beta_loc);

cfg = decoding_describe_data(cfg,labelnames,labels,regressor_names,beta_loc);
  
cfg.design = make_design_similarity_cv(cfg); 
  
results = decoding(cfg);

Any suggestions on how to specify/adjust the correct averaging procedure?

Thanks in advance!

Ricardo.