Individual size distributions (ISD; also known as size spectra, community biomass distributions) are a powerful framework for understanding biological communities by incorporating processes from the individual- to the population-level. ISD relationships are described by a power law: N ~ Mλ, where N is abundance and M is individual body mass. Because M is a continuous measure, the relationship simplifies to a frequency distribution in the form of f(M) ~ Mλ. Determining the value of λ from a vector of individual body sizes, m, is a frequent goal of ecologists. However, typically have biases associated with them, particularly with under-sampling of the smaller values. These empirical data biases result in only some portion of the right tail of the data fitting a power law distribution well. Although the potential issues of estimating λ from a biased empirical sample have been previously discussed, and solutions (typically “cutting” the data below some arbitrary size) have been offered, to our knowledge no formal assessment of the effects that biased data has on estimates of λ have been performed. Here, we used a numerical simulation study design to assess how biases (specifically, under-sampling of small values) affect estimates of λ, as well as testing a solution proposed by Clauset et al. (2009) for determining the best point to “cull” the data to maximize the fit to a power law distribution. We found that estimates of λ became less accurate as the degree of under-sampling increased. When the under-sampled data was culled, λ estimates universally improved, although the improvement was not as great with the extremely under-sampled data. Based on our results, we recommend that all future studies of individual size distributions pre-process their data to remove biases associated with under representation of the smaller values. This will improve estimates of λ and will aid in understanding the remarkable consistency of λ across biological communities, and, importantly, to better understand when and why λ does vary.