Title: Mitigating the statistical bias due to anatomical variation in pediatric fNIRS Abstract: Functional near-infrared spectroscopy (fNIRS) is a non-invasive neuroimaging method that records changes in blood oxygenation in the brain using optical sensors placed on the scalp. Although this technique is widely used for functional imaging, this approach does not provide any direct anatomical information about the structure of the head, such as layer thicknesses and brain depth beneath the sensors. As a result, the magnitude of fNIRS signals is confounded by variations in this unknown underlying anatomy which introduces the potential for statistical bias when comparing across groups and in longitudinal studies. This is particularly relevant for pediatric and developmental studies where this anatomy is expected to systematically vary across sample groups. Currently in the fNIRS field, corrections for this unknown anatomy have been proposed using semi-empirical formulas, look-up tables, or off-the-shelf statistical mediation models such as mixed-effects modeling. However, these corrections are not able to account for the complexity of this problem and unable to model heterogeneity across subjects and spatial locations. In this work, we propose to develop a statistical correction method based on the construction of a database of distributions of these corrections across demographics and spatial locations on the head obtained from analysis of existing pediatric anatomical MRI data. We propose to use these distributions as priors in the analysis of fNIRS group level data to model bias and uncertainty introduced by unknown anatomical structures. Aim 1. Characterize how key anatomical and demographic factors influence fNIRS measurements. Aim 2. Development and characterization of our proposed novel anatomical statistical distribution model. Aim 3. Integrate the proposed model into our existing open source fNIRS toolbox.