Hi everyone, I'm new here and getting my first post in with a problem.
Im trying to implement a strategy from this paper: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1509811
This is mainly a learning exercise for myself. Basicly the algo applies an eigenvalue filter to identify if the market is good to trade on. I have implemented this part, and am up to 2c) in the algo as identified in the paper.
Im struggling to get my head around the state vectors and there uses. If i write out how I interpret the steps, I would be beyond grateful if somebody with a bit more experience could point out if I am wrong! I understand that this would probably not be the exact implementation (you could calculate some things at different times to make things more seamless im sure).
2c) find the various moments of the vector of returns for each asset. the return vector is the form r(1) = return at t-tau+1, r(2) = return at t-tau+2 etc. so this step would give a single mean value for each moment and asset. where t is the most recent time, and the return is that given by the return matrix.
2d) average each moment for all assetts. so this gives four values, the mean return, variance, scew and kurt over the past tau days.
2e) calculate again, but rather than t as the latest time, repeat this for all values of t from tau+2 to latest-1. use all previous values to find the percentile each moment sits in (high, mid, low). For simplicity, state low would 1, mid 2 etc.
then save the average of each moment in a state vector, so a high mean, low var, med scew, med kurt would be added to the vector denoted by 3,1,2,2.
2f) for the most recent stats calculated in d), find the percentile of each one using all values calculated in e), and calculate the state vector for the most recent readings. for example, the latest one may be 1,2,3,2.
2g) go into the vector 1,2,3,2 which was created by calculating 2e). If this is blank invest in risk free, and add the current readings. if this has several values, eg. the mean returns are 0.5 1 0.5 2, and variances 2,2,2,2, average each moment, and then calculate mean over variance and apply the rules to make a decision of investment. so in this example it would be 1/2, meaning invest in risk free.
Im not 100% sure where the logaritihmic return kv is used, if anybody knows I would be most grateful!
I hope that was clear, please ask if anything isnt!
thanks Chris
Im trying to implement a strategy from this paper: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1509811
This is mainly a learning exercise for myself. Basicly the algo applies an eigenvalue filter to identify if the market is good to trade on. I have implemented this part, and am up to 2c) in the algo as identified in the paper.
Im struggling to get my head around the state vectors and there uses. If i write out how I interpret the steps, I would be beyond grateful if somebody with a bit more experience could point out if I am wrong! I understand that this would probably not be the exact implementation (you could calculate some things at different times to make things more seamless im sure).
2c) find the various moments of the vector of returns for each asset. the return vector is the form r(1) = return at t-tau+1, r(2) = return at t-tau+2 etc. so this step would give a single mean value for each moment and asset. where t is the most recent time, and the return is that given by the return matrix.
2d) average each moment for all assetts. so this gives four values, the mean return, variance, scew and kurt over the past tau days.
2e) calculate again, but rather than t as the latest time, repeat this for all values of t from tau+2 to latest-1. use all previous values to find the percentile each moment sits in (high, mid, low). For simplicity, state low would 1, mid 2 etc.
then save the average of each moment in a state vector, so a high mean, low var, med scew, med kurt would be added to the vector denoted by 3,1,2,2.
2f) for the most recent stats calculated in d), find the percentile of each one using all values calculated in e), and calculate the state vector for the most recent readings. for example, the latest one may be 1,2,3,2.
2g) go into the vector 1,2,3,2 which was created by calculating 2e). If this is blank invest in risk free, and add the current readings. if this has several values, eg. the mean returns are 0.5 1 0.5 2, and variances 2,2,2,2, average each moment, and then calculate mean over variance and apply the rules to make a decision of investment. so in this example it would be 1/2, meaning invest in risk free.
Im not 100% sure where the logaritihmic return kv is used, if anybody knows I would be most grateful!
I hope that was clear, please ask if anything isnt!
thanks Chris
