Hotspots are widely used to construct a reference frame to track the motion of plates relative to lower mantle. How fast hotspot move is crucial in constructing such reference frame. In addition, hotspot motion itself bears some implication about mantle dynamics. While previous studies show different opinion about hotspot motion. Slow motion is supported by many studies, such as Morgan [1981, 1983] and Duncan [1981], which estimated that individual hotspots move relative to a mean hotspot reference frame by 3-5 mm a-1. In contrast, more recently, it has been popular to assume that hotspot in the Pacific basin move relative to those in the Indo-Atlantic at rates of 10-30 mm a-1 or more [Norton, 1995; Raymond et al., 2000; Tarduno et al., 2003]. And fast motion motivates the usage of a moving hotspot reference (e.g. Doubrovine et al., [2012]) with modeled hotspot motion. This dissertation aims to evaluation hotspot motion and hotspot reference frame using an updated global hotspot data set (HS4, revised from by Morgan and Phipps Morgan [2007]). In the first part, I fit hotspot trends from HS4 and calculate hotspot motion from data misfit. It shows that the angular misfit of fitting observed hotspot trends is dominated by differences in trends between plates rather than difference within plates, which indicate misfit of fitting hotspot trends in the same plate are correlated with each other. And the estimated plate-mean hotspot motion in the direction perpendicular to plate motion differ significantly from zero for hotspot beneath only 3 of 10 plates and then only by 0.3 to 1.4 mm a-1. In the second part, we objectively estimate uncertainty in hotspot trend using hotspot track width and length for hotspot in HS4 data set. It shows that uncertainties in trends are underestimated by most previous studies [e.g. Morgan and Phipps Morgan, 2007]. Our newly estimated uncertainties in trends are consistent with the misfits observed in the last part. We update plate-mean hotspot motion using the new uncertainties in trends, 54 of 56 hotspot motion differ insignificantly from zero and none of 10 plates differ significantly from zero. In the third part, we test hotspot fixity by fitting 41 hotspot trends (T41, a subset of HS4) using both fixed hotspot and moving hotspot reference frame. The modeled hotspot motions used in moving hotspot reference frame are calculated from Doubrovine et al. [2012]. The results show that more hotspots and more plates are fitted better in fixed hotspot reference frame than moving hotspot reference frame. Assuming modeled hotspot motion directions are correct, we calculate an average hotspot motion for T41 is 1.5 mm a-1, with lower bound zero and upper bound as 4.6 mm a-1 for magnitude (95% confidence limit). Thus, fixed hotspot reference frame is superior than moving hotspot for T41, and possibly for all available hotspot data set. In the fourth part, we calculate a new of absolute plate motion using both hotspot trend and orientation of seismic anisotropy (HS4-SKS-MORVEL). We first calculate a new set of absolute plate motion using only hotspot trends data. (HS4-MORVEL). HS4-MORVEL is superior than previous plate motions using hotspots. First, it considers that hotspots in the same plate are correlated with each other by using a two-tier analysis. Second, it uses newly estimate hotspot trend uncertainties. Then we combine HS4-MORVEL and SKS-MORVEL [Zheng et al., 2014] to get HS4-SKS-MORVEL, which has better resolution than both sets.