copulafit.m
function  nll = negloglike_t(nu, R, t)  
% Compute negative log-likelihood for a t copula at nu and R = chol(Rho)
    [n,d] = size(t);
    % R = R ./ repmat(sqrt(sum(R.^2,1)),d,1);
    R = bsxfun(@rdivide, R, sqrt(sum(R.^2,1)));

    % nll = -sum(log(mvtpdf(t,R'*R,nu)) - sum(log(tpdf(t,nu)),2)),
    % where t = tinvLocal(u,nu)
    tRinv = t / R;
    nll = - n*gammaln((nu+d)/2) + n*d*gammaln((nu+1)/2) - n*(d-1)*gammaln(nu/2) ...
          + n*sum(log(abs(diag(R)))) ...
          + ((nu+d)/2)*sum(log(1+sum(tRinv.^2, 2)./nu)) ...
          - ((nu+1)/2)*sum(sum(log(1+t.^2./nu),2),1);
end

建议用copulafit函数试试