Average of several (different, across) structures

I have 10 structures with 6 fields, each with 3D arrays. All of them have the same organization.

I want to do the average of the 10 structures, as a result I would have 1 mean structure with 6 fields, each with 3D arrays.

For example, I have:

Subject_1_RMSE_angle = 
  struct with fields:
    narxwalk: [17×10×16 double]
     EMGwalk: [17×10×16 double]
      narxup: [17×10×16 double]
       EMGup: [17×10×16 double]
    narxdown: [17×10×16 double]
     EMGdown: [17×10×16 double]
     
 Subject_2_RMSE_angle = 
  struct with fields:
    narxwalk: [17×10×16 double]
     EMGwalk: [17×10×16 double]
      narxup: [17×10×16 double]
       EMGup: [17×10×16 double]
    narxdown: [17×10×16 double]
     EMGdown: [17×10×16 double]
     
  Subject_3_RMSE_angle = 
  struct with fields:
    narxwalk: [17×10×16 double]
     EMGwalk: [17×10×16 double]
      narxup: [17×10×16 double]
       EMGup: [17×10×16 double]
    narxdown: [17×10×16 double]
     EMGdown: [17×10×16 double]

then do the mean, and have a results as:

Subject_MEANs_RMSE_angle = 
  struct with fields:
    narxwalk: [17×10×16 double]
     EMGwalk: [17×10×16 double]
      narxup: [17×10×16 double]
       EMGup: [17×10×16 double]
    narxdown: [17×10×16 double]
     EMGdown: [17×10×16 double]

Thank you.

[MARCO BERGONZI]

function [outmode,outmean,outmedian,outprcile]=mfromstruct(nomestruttura,nomecampo,percentile_voluto)

pos=length(nomestruttura);

x=zeros(pos,1);

t=numel(nomestruttura(1).(nomecampo));

outmode=nomestruttura(1).(nomecampo)*0;

outmean=outmode;

outmedian=outmode;

outprcile=outmode;

for j=1:t

for i=1:pos

x(i)=nomestruttura(i).(nomecampo)(j);

end

outmode(j)=mode(x);

outmean(j)=mean(x);

outmedian(j)=median(x);

outprcile(j)=prctile(x,percentile_voluto);

end

end

Answer 1

[Stephen23]

Fake data in scalar structures:

S1 = struct('A',randi(99,4,3,2),'B',randi(99,4,3,2));
S2 = struct('A',randi(99,4,3,2),'B',randi(99,4,3,2));
S3 = struct('A',randi(99,4,3,2),'B',randi(99,4,3,2));

Concatenate into one non-scalar structure (which they should be anyway):

SX = [S1,S2,S3];

Concatenate data from each field, calculate mean:

C = fieldnames(SX);
SZ = struct(); % scalar structure
for k = 1:numel(C)
    F = C{k};
    SZ.(F) = mean(cat(4,SX.(F)),4);
end

SZ is a scalar output structure containing the means of the input scalar structures.