Background. Allometric scaling (AS) is widely used in predicting human clearance (CL) based on animal data. Substantial prediction errors have been commonly observed and various modifications to AS have not provided a broad reliable improvement. In this study, an extensive data set was assembled including animal and human systemic CL and physiochemical properties. The allometric exponents were calculated based on multiple species AS and single-species AS methods. The correlations between the allometic exponents and physiochemical properties were evaluated in an attempt to find covariates that may explain the inter-compound variability in the allometric exponents. Lastly, the statistical approaches in analyzing the allometric function were evaluated with the collected data. Methods. 1- A nonlinear mixed effect modeling (MEM) approach was performed to investigate the central tendency and distribution of AS exponents as well as to identify whether there are any correlations between the allometric exponent, and coefficient, with the physicochemical and drug metabolism and pharmacokinetics (DMPK) properties of the compounds. 2- Single-species AS was performed to estimate the single-species AS exponent distributions and their corresponding central tendencies. The correlation between the estimated single-species AS exponents and the physicochemical and DMPK properties of the compounds were also examined. 3- The methodologies of log-log transformation followed by linear regression (LL-LR) and direct nonlinear regression methods (NLS) with different weighting schemes on the AS power function were investigated. The central tendency and distribution of the allometric exponents were evaluated and compared across methods. Furthermore, the human CL prediction performance was evaluated among methods. Results. The estimated central tendency and distribution of AS exponents from the nonlinear MEM as well as the single-species AS approaches were consistent with literature reports. There were no significant correlations identified between the estimated AS exponents and the physicochemical or DMPK properties. The methods of LL-LR and the NLS with 1/w2 weighting (variance weighted by CL2 during the variance minimization process) results in the most similar allometric exponent with central tendency around 0.668 and provided the best human CL prediction among methods investigated. Conclusion. The knowledge gained in this work by extensive modeling and simulations contributed to a better understanding of the variability in AS exponents and better practice in performing AS in human CL prediction