Spearman Rank Decay

[TurtleDaddy]

Hello everyone!

This may be a bit irrelevant to the FRM examination, but i came across this Credit Suisse research pdf (http://www.scribd.com/doc/62210581/CSFB-Quant-Research) and was wondering about the implementation of spearman rank decay analysis presented on pp.27.

If we take exhibit 34 as an example, what have the authors done? Have they:

1. Ranked the universe of stocks in terms of the Traditional Value factor at time t ; Ranked the universe of stocks in terms of returns at time t+1 and computed the rank correlation qnd repeated this step for future months.(E.g. Rank stocks in terms of factor in Jan;Rank stocks in terms of future returns in Feb;compute rank correlation)
2. OR Ranked the universe of stocks in terms of the Traditional Value factor at time t;Ranked the universe of stocks in terms of returns at time t+1;computed the rank correlation statistic; and extended the data set to include more periods before calculating the rank again. (E.g. Rank stocks in terms of factor in Jan;Rank stocks in terms of returns in Feb;calculate rank correlation;extend data set from jan to feb for the factor and from feb to march for returns;recalculate rank correlation)
3. OR something else completely?

Any help would be much appreciated!

 

[Bryon]

Hi,
If it could be of any help, there is a section on Spearman rank correlation testing in CFA level 1 quant book.
You can find the procedure, t-test formula and distribution table to test correlation.
In short, let X and Y 2 series to be tested.
1. Rank X from largest to smallest, assign rank 1,2 etc to Xi from largest=1. Do the same for Y.
2. Calculate the rank differences di of each pair of observation (Xi,Yi)
3. Calculate Spearman rank correlation Rs which is a function of di and sample size n
4. Hypothesize your hypothesis
5. Calculate the t-test ( a function of Rs and n)
6. Check against look-up table to see if you could reject the null hypothesis.