This work describes an experimental study to assess if analytical and empirical models can estimate the risk of warping deviation for parts made using fused filament fabrication based on part geometry. We also examine how the accuracy of the prediction varies for different machines and materials. If the predictive models can estimate risk of warping for a given part geometry, they can help enable better design for additive manufacturing so that designers can change part geometry early in the design process to have more easily-manufacturable parts, or choose an alternative orientation to optimize dimensional accuracy at the process planning stage. Specifically, we evaluate the extent to which two analytical models and one empirical model can assess the risk of warping for approximately rectangular parts with varying dimensions. We analyze dimensional accuracy data for parts with different length, height, and fillet type that were printed in ABS and PLA on different fused filament fabrication machines. After evaluating the three models, we found that the empirical model had the best performance over all datapoints. However, the analytical models showed promise but need further refinement on how the prediction of warping deviation depends on part height. Areas for additional research are highlighted.