I want to average multiple NARX networks on the same set of input-targets: Am I doing it right?

Here's the relevant chunk of code:

numNN = 100;

j=1; % Prepare average cell arrays [Xc,Xic{j},Aic{j},Tc] = preparets(netc,X,{},T);

Xic{j}={0}; Aic{j}={0};

[Xict_,iX,jX] = nncell2mat(Xic{j}); [Aict_,iA,jA] = nncell2mat(Aic{j});

j=0; for i = 1:numNN

    % Prepare the Network
    [Xc,Xic{i},Aic{i},Tc] = preparets(netc,X,{},T);
    % Train the Network
    netf = train(netc,Xc,Tc,Xic{i},Aic{i});

    Yc = netf(Xc,Xic{i},Aic{i});
    closedLoopPerformance = perform(netf,Tc,Yc);
    tc = cell2mat(Tc); 
    vartc = mean(var(tc',1));
    Ec = gsubtract(Tc,Yc); 
    NMSEc = mse(Ec)/vartc

    [Xic_,iX,jX] = nncell2mat(Xic{i});
    [Aic_,iA,jA] = nncell2mat(Aic{i});

    if(NMSEc<2.0e-3)
        j=j+1;
        Xict_= Xict_ + Xic_;
        Aict_= Aict_ + Aic_;
    end
    [i j]
end

Xict_=Xict_/j; Aict_=Aict_/j;

Xfc = mat2cell(Xict_,iX,jX); Afc = mat2cell(Aict_,iA,jA);

% Multi-step Prediction

X2 = X1(1, 1); Y2 = netf(X2,Xfc,Afc); Ypred(1)=Y2; % % % %

Any input is appreciated, thank in advance for your time. Please let me know if you need any additional information.