Association between monthly leptospirosis cases and El Niño and meteorological variables, controlling for 12-month harmonic variation.

Selected single predictor models. Smaller values of BIC signify a better fit to the data. All lags between 0 and 6 months were tested for each variable. For the cumulative sum variables, all ranges from the previous 0 to 9 months were tested. Only the lag or cumulative sum with the lowest BIC score for each variable is presented here. A difference in the BIC scores of greater than 2 is considered to be important. All of the models include sine and cosine terms with a 12-month period. For the full table of results, see Supporting Table S3.

经筛选的单预测变量模型。贝叶斯信息准则（Bayesian Information Criterion，BIC）取值越小，代表模型对数据的拟合效果越好。针对每个变量，我们测试了0至6个月的所有滞后阶数；对于累积和变量，则测试了此前0至9个月的全部区间范围。本文仅展示每个变量对应最低BIC得分的滞后阶数或累积和结果。当BIC得分差值大于2时，可认为该差异具有重要意义。所有模型均包含周期为12个月的正弦项与余弦项。完整结果表格详见补充表S3。