开发者:上海品职教育科技有限公司 隐私政策详情

应用版本:4.2.11(IOS)|3.2.5(安卓)APP下载

Jingwen · 2023年07月13日

老师能不能解释一下look-ahead bias

NO.PZ2023010903000045

问题如下:

Leeter tells Clickman about the interesting investment process of the XTZZ Fund. Leeter states that many times, government- reported data are revised three to six months after the data are initially reported. These revisions then become incorporated into the historical data, with the revised value replacing the originally reported value. When the revision is incorporated, the data are referred to as “clean data.” XTZZ applies its analysis in a manner in which the originally reported data exist as a lagged factor until the revised data become available. This approach differentiates XTZZ from funds that use only the “clean” data, which ignore the initially reported data values.

XTZZ’s approach to analyzing government-reported data most likely reduces:

选项:

A.

look-ahead bias

B.

survivorship bias

C.

model overfitting

解释:

XTZZ’s approach prevents the use of “revised” government data that are not known when the data are initially reported. By not incorporating revised government data until they are actually revealed, XTZZ reduces look-ahead bias.

B is incorrect. Survivorship bias occurs when analysis on past data ignores firms that have ceased to exist at some point during the analysis period.

C is incorrect. Model overfitting occurs when the model fits the past data by finding a pre-conceived relationship (i.e., making the model specifications biased toward the pre-relationship) or when the model is biased toward patterns that are specific only to the past data.

是数据有滞后性,就会有look-ahead bias吗?还是数据太前沿了就不对?

麻烦老师解释一下,谢谢。

1 个答案
已采纳答案

笛子_品职助教 · 2023年07月14日

嗨,努力学习的PZer你好:


look-ahead bias


Hello,亲爱的同学~

look - ahead bias是指:用了当时未公布的数据。


比如公司年报会在4月1日公布,但公司年报是描述的12月31日的经营情况。

如果在历史回测的时候,在1月1日就使用了这个年报里的数据,就是look ahead bias.


再比如,GDP数据会先发布一个初始数据,比如4月1日公布一季度的GDP增长率为5%。

然后在年末会对数据进行复查,如果有必要,会发布一个修正数据,比如12月31日发布,一季度GDP数据不是初始公布的5%,而是4%。

然后策略,在4月1日,使用了一季度GDP = 4%这个数据,这就是look - ahead bias。因为4月1日并不知道4%这个数据。


国内也有类似的概念,在量化里有有一个未来函数的术语,就是“未来函数”就是“Look - ahead bias”.


百度上解释如下:

----------------------------------------------
加油吧,让我们一起遇见更好的自己!

  • 1

    回答
  • 0

    关注
  • 457

    浏览
相关问题

NO.PZ2023010903000045 问题如下 Leeter tells Clickmabout the interesting investment process of the XTZZ Fun Leeter states thmany times, government- reporteta are revisethree to six months after the ta are initially reporte These revisions then become incorporateinto the historictwith the revisevalue replacing the originally reportevalue. When the revision is incorporate the ta are referreto “cleta.” XTZZ applies its analysis in a manner in whithe originally reporteta exist a laggefactor until the reviseta become available. This approafferentiates XTZZ from fun thuse only the “clean” twhiignore the initially reporteta values.XTZZ’s approato analyzing government-reporteta most likely reces: A.look-ahebi B.survivorship bi C.mol overfitting XTZZ’s approaprevents the use of “revise government ta thare not known when the ta are initially reporte not incorporating revisegovernment ta until they are actually reveale XTZZ reces look-ahebias.B is incorrect. Survivorship bioccurs when analysis on past ta ignores firms thhave ceaseto exist some point ring the analysis perioC is incorrect. Mol overfitting occurs when the mol fits the past ta finng a pre-conceiverelationship (i.e., making the mol specifications biasetowarthe pre-relationship) or when the mol is biasetowarpatterns thare specific only to the past t 老师,可以再一下overfitting bias吗

2024-07-02 21:23 1 · 回答

NO.PZ2023010903000045问题如下 Leeter tells Clickmabout the interesting investment process of the XTZZ Fun Leeter states thmany times, government- reporteta are revisethree to six months after the ta are initially reporte These revisions then become incorporateinto the historictwith the revisevalue replacing the originally reportevalue. When the revision is incorporate the ta are referreto “cleta.” XTZZ applies its analysis in a manner in whithe originally reporteta exist a laggefactor until the reviseta become available. This approafferentiates XTZZ from fun thuse only the “clean” twhiignore the initially reporteta values.XTZZ’s approato analyzing government-reporteta most likely reces: A.look-ahebiasB.survivorship biasC.mol overfitting XTZZ’s approaprevents the use of “revise government ta thare not known when the ta are initially reporte not incorporating revisegovernment ta until they are actually reveale XTZZ reces look-ahebias.B is incorrect. Survivorship bioccurs when analysis on past ta ignores firms thhave ceaseto exist some point ring the analysis perioC is incorrect. Mol overfitting occurs when the mol fits the past ta finng a pre-conceiverelationship (i.e., making the mol specifications biasetowarthe pre-relationship) or when the mol is biasetowarpatterns thare specific only to the past txtzz applies….这句话能翻一下吗

2023-07-19 21:11 1 · 回答