// low-pass filter design
  const unsigned int len = 20; // number of filter coeffs/tabs
  const ScalarType cutoff = 0.1; // normalized cutoff frequency
  const ScalarType lag = 0; // filter lag 0 implies "zero" lag, i.e. real time filtering
  const ScalarType lambda = 0; // lambda emphasizes timeliness vs. accurateness, 0 implies a the MSE solution, try to increase it!
  const ScalarType eta = 0; // eta emphasizes smoothing vs. accurateness, 0 implies a the MSE solution try to increase it!
  const Vector per = DFA::periodogram(data); // compute the periodogram, i.e. an estimate of the signal's power spectry density
  const DFA::FIR fir = DFA::designLP(per, len, cutoff, lag, lambda, eta); // design the low-pass filter

  // offline filtering
  const Vector offlineSignal = DFA::apply(data, fir.coefficients);

  // online filtering
  Vector onlineSignal = Vector::Zero(data.size());
  DFA::OnlineFIR filt = DFA::OnlineFIR(fir.coefficients);
  for (unsigned int i = 0; i < data.size(); ++i)
    onlineSignal[i] = filt.process(data[i]);