Dataset description: Accompanying data for manuscript “Multi-Site, Multi-Vendor Validation of the Accuracy and Reproducibility of Proton-Density Fat-Fraction Quantification at 1.5T and 3T using a Fat-Water Phantom”, submitted for publication in Magnetic Resonance in Medicine in January 2016. Details: The 'datasets' folder contains multiple MATLAB MAT-files files with chemical shift-encoded (CSE) MRI data for validation of fat quantification techniques in a fat-water phantom. Data was acquired at six sites with different MRI vendors, two field strengths (1.5T and 3T) per site, and two protocols per field strength. The goal of these techniques is to measure proton-density fat-fraction (PDFF) accurately and reproducibly. Data from site 1 was acquired both at the beginning (October 2014) and at the end (December 2015) of this study, in order to evaluate the integrity of the phantom. Details on the vendors, platforms and protocols are described in the file 'vendors_platforms_protocols.pdf' in this same folder. The phantom consists of 11 vials with different fat concentrations (PDFF = 0%, 2.6%, 5.3%, 7.9%, 10.5%, 15.7%, 20.9%, 31.2%, 20.9%, 31.2%, 41.3%, 51.4% and 100%, respectively). At each magnet, the vials were arranged horizontally along the B0 field and scanned with axial slices using a 3D multi-echo spoiled gradient echo pulse sequence. Each of the MAT-files contains the acquired complex-valued images over the three central slices within the phantom vials. The structure 'imDataAll' contains the acquired data, with the following fields: 'TE' (echo times), 'FieldStrength' (in Tesla), 'PrecessionIsClockwise' (describing whether water has higher resonance frequency than fat according to the reconstruction convention), 'echo_polarity' (relative polarity of the acquired echoes, ie: monopolar vs bipolar readouts), 'images' (five-dimensional array containing the complex valued images, coil-combined in datasets received with multiple channels, with dimensions X x Y x Slices x Coils x Echoes). Additionally, the MAT-files contain reconstruction results as described in the manuscript, in the arrays 'fwmc_*' of size X x Y x Slices. Note that the nomenclature 'fwmc' stands for 'Fat-Water separated with Magnitude fitting (performed after complex fitting in order to obtain full 0-100% range of PDFF while avoiding errors related to phase shifts in the data), and a Common initial phase for the water and fat signals. The specific arrays are 'fwmc_ff' (PDFF map), 'fwmc_r2star' (R2*=1/T2* decay rate), 'fwmc_w' (water image), 'fwmc_f' (fat image).