survivorship/look-ahead bias criteria, discounted, remember, selected, question

waldziuchna

I went with Survivorship bias too

spartan1

There was a question about a model based on a random selection of stocks, their performance over the past maybe 20 years. Two of the biases offered were survivorship and look-ahead, don't remember the third, but discounted it quickly in the exam. I hesitated between these two, because I felt the question was crying out for "look-ahead", but saw nothing to suggest that the sample was selected based on criteria subsequently revealed.

I went for survivorship on the basis that the performance of a random selection of stocks from the present (which was my understanding) would not include the data of stocks from the earlier period since extinct. I'm uneasy, though, about the answer, as I remember feeling it wasn't clear-cut.

Anyone any thoughts on this?

dmar

I went for look ahead bias

Wasteoftime

the other option was data mining bias.

koba

If the third choice was data mining, so the right answer could be this one... because the definition of data mining is : significant relationships that have occured by chance. If we assume that 20 years is not suficiently long to be considered valid.

JRossSter

I said look ahead. Looks like we're all over this one.

ogoluwa

This was a tricky question... don't remember exactly why but after spending extra time digging into the wording and eliminated Look-ahead and Survivorship. I guess we'll never know.



Edited 1 time(s). Last edit at Thursday, June 25, 2009 at 10:59AM by bid_offer_shark.

joemoran

i went with data mining too

rgonzalez

I did not remember the 3rd choise when I responded to lmb..

I think I answered Data Mining ... looks right to me.

btcapital

I think I quickly discounted data-mining because there was nothing necessarily narrow or derived in the sample selection, and no particular criteria that were being "over-worked".