MLE fits to step lengths and normalized step lengths (N = 145,731 steps).

Negative log-likelihood measures the relative ability of candidate models to explain the observed data (For additional fits tested, see S1 and S2 Tables). The corrected Akaike information criterion (AICc) and Bayesian information criterion (BIC) (S2 Table) confirm that order of fit quality is not due to the number of model parameters. The most negative log likelihood and AICc scores are the best fits; in this case that is the smallest positive score for the lognormal distribution. The last column lists the distribution parameters that were selected by MLE. See S1 and S2 Tables for other distribution fits and goodness of fit statistics.

负对数似然（Negative log-likelihood）用于衡量候选模型解释观测数据的相对能力（如需查阅其他经检验的拟合模型，详见附表S1与S2）。修正的赤池信息准则（corrected Akaike information criterion, AICc）与贝叶斯信息准则（Bayesian information criterion, BIC）（详见附表S2）证实，拟合质量的优劣并非由模型参数数量所决定。负对数似然与AICc得分越低，模型拟合效果越好；本案例中，对数正态分布的负对数似然得分即为最小的正值，对应最优拟合结果。最后一列列出了通过极大似然估计（Maximum Likelihood Estimation, MLE）筛选得到的分布参数。如需了解其他分布的拟合结果与拟合优度统计量，详见附表S1与S2。