rdm_neurom.m

function [mods orthmods] = rdm_neurom(kap,off,gnorm);
%(c) Bernhard Spitzer, 2016

%% %%%%%%%%%%%%%%%%%%%% psychometric models %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

f=[-1:0.4:1];
psym=abs((off+f).^kap).*sign(off+f);
if gnorm
    g=sum(abs(f+off).^kap) / sum(abs(f));
    psym=psym./g;
end
psym=[-psym;psym];

%% %%%%%%%%%%%%%%%%%%%% theoretical models %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% in here only for orthogonalization
%%  digit
pc=ones(6,6);
for i=1:6
    for j=1:6
        if i==j  % remove diagonal (6x6)
            pc(i,j)=0;
        end
    end
end
c=[pc pc; pc pc];
mods.dig=c;

%% category
c=[zeros(6,6) ones(6,6); ones(6,6) zeros(6,6)];
for i=1:12
    for j=1:12
        if i==j  % remove diagonal
            c(i,j)=0;
        end
    end
end
mods.cat=c;

%%  parity
v1=repmat([0 1],1,6);
v2=repmat([1 0],1,6);
c=repmat([v1;v2],6,1);
mods.eve=c;

%% %%%%%%%%%%%%%%%%%%%% behavioral models %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% signed weight (~ psychom numerical distance)
p=[-psym(1,:) psym(2,:)];
c=[];
for i=1:12
    for j=1:12
        c(i,j)=abs(p(i)-p(j));
    end
end
squtest=squareform(c);
mods.numd=c;

%% response-mapped weight
p=[psym(1,:) psym(2,:)];
c=[];
for i=1:12
    for j=1:12
        c(i,j)=abs(p(i)-p(j));
    end
end
squtest=squareform(c);
mods.nXc=c;

%% orthogonalize models
models=fieldnames(mods);
vecs=[];
for m=1:length(models)
    vecs(:,m)=squareform(mods.(models{m}));
    vecs(:,m)=vecs(:,m)-mean(vecs(:,m)); % mean center
end

for m=1:length(models) % recursive orthogonalization
    order4orth=[find((1:length(models))~=m) m];
    tmpvecs=spm_orth(vecs(:,order4orth));
    orthmods.(models{m})=squareform(tmpvecs(:,end));
end

% %% plot model RDMs
% figure; colormap('hot');
% for m=1:length(models)   
%     subplot(2,length(models),m); 
%     imagesc(mods.(models{m}),[0 1]); title(models{m});    
%     subplot(2,length(models),m+length(models));
%     imagesc(orthmods.(models{m}),[-1 1]); title([models{m} '_{orth}']);
% end