How random measurement errors propagate to the chromophore concentration calculation in broadband near-infrared spectroscopy