Background: Respiratory motion (RM) can degrade image quality in PET. Various methods had been proposed to correct RM. The standardized uptake value (SUV) is often used as the figure of merit for evaluating the correction efficacy. However, SUV is not an objective metric to measure RM correction (RMC) effectiveness for the following reasons: 1) SUV will not identify variations in correction effectiveness for different respiratory motion subtypes, e.g., large variation in motion amplitude and baseline drifting among respiratory cycles; 2) attenuation correction, scatter correction and parameters used in the reconstruction, will affect accuracy of SUV, independent of the quality of RMC. For example, AC artifacts can introduce SUV mis-estimation due to an RM introduced mismatch between PET and AC map; 3) SUV depends on the region of interest (ROI) definition, which is often manually drawn and can therefore be subjective. Methods: We propose a new method, i.e., motion-corrected centroid-of-distribution (MCCOD), to evaluate the efficacy of a RM model using the TOF PET list-mode data only. The model performance is summarized by a single scalar, i.e., respiratory signal to noise ratio (rSNR). We demonstrate how to use MCCOD to evaluate two types of RM models: 1) inter-gate registration model which is typically used in a motion-corrected image reconstruction (MCIR) method and 2) internal-external motion correlation (INTEX) model, which is used for continuous motion correction. Two types of gating methods, i.e., amplitude-based (AG) and phase gating (PG) were also evaluated. Results: MCCOD was shown to be effective in evaluating the correction effectiveness. Motion models built based on AG outperformed models with PG. Based the same gating method, INTEX motion models outperform MCIR models. We will use this method to compare with image-based metrics, e.g., SUV.