1. 是每一个公司 (StockCode) 都要做吧？ 2. 以 StockCode=39 为例，这么多不同时间的预测，我怎知不会有同一分析师在 90 天内重复预测呢？

此外，资料请给多一点，例如 100 笔。

----------------------- copy starting from the next line -----------------------参考文献是这样说的the median analyst forecast using only analyst
forecasts announced in the 30 calendar days preceding the analyst forecast.
另有附加信息是To facilitate an apples-to-apples comparison of management and analyst forecast news, we restrict the sample to management forecasts and the corresponding analyst forecasts for the same firm that have the same fiscal period end and that are for the same periodicity (annual vs. quarterly). We also create a comparable measure of analyst forecast news by defining News for analyst forecasts as the difference be-
tween an individual analyst’s forecast and the consensus analyst forecast,all scaled by the absolute value of this consensus forecast. When there is more than one analyst forecast made for the same firm on the same day for the same fiscal period end and periodicity, we calculate the mean value of News across the analysts.我按照附加信息以管理层盈余预测所属会计期间匹配了对应的分析师预测，样本量就缩减了很多，我的想法是要使每个分析师盈余预测发布日对应一个完整的分析师分布日周期，在相减，相同公司相同分析师的也可以相互相减  只要相差的日期满足90天（英文文献取得30天，我自己做的90天）

你似乎没回答上面 (至少) 第二个问题！

请 ssc install rangestat 并试试