Integrating noisy data

Suppose the parametric form of a curve is not known, but only a set of observations. Quadrature formulae can be used to integrate a function only known from a set of data points. However, the results will be unreliable if the data contains measurement errors (noise). The method presented here fits an even degree piecewise polynomial to the data where all the data points are being used as knot points and the smoothing parameter is optimal for the indefinite integral of the curve which happens to be a smoothing spline. After the smoothing parameter has been chosen, this approach is less computationally expensive than fitting a smoothing spline and integrating.