This is truly a very useful article, not just for reading but to put in practice.
I have a question regarding the Cornish Fisher Transform. You talk about calculating rolling skew, kurtosis, but don't see that in the code.
The transform should be applied on a rolling basis, right? Otherwise, this will create a look-ahead bias. or is this supposed to be applied like a StandardScaler for preprocessing?
actually would also be great to see an example application of this class. of course we can figure it out, but I think it'll enhance the article. just a suggestion.
I have a question-- before orthogonalization, would you align all assets to the same time points by using each asset’s most recently completed bar, so every asset ends up with the same number of rows even though volume bars occur at different times?
suppose I have a futures series, instead of stock. the series is roll adjusted (by ratio), so the past prices are not historical prices, but they preserve the ratio p(t+1)/p(t). Would the fractional differencing work in that case?
Fractional diff can work on a roll-adjusted futures series, but you should apply it to the log series, not the raw adjusted price level. It works only for research. For production it is a pitfall unless the series is constructed point-in-time
This is truly a very useful article, not just for reading but to put in practice.
I have a question regarding the Cornish Fisher Transform. You talk about calculating rolling skew, kurtosis, but don't see that in the code.
The transform should be applied on a rolling basis, right? Otherwise, this will create a look-ahead bias. or is this supposed to be applied like a StandardScaler for preprocessing?
Hi Dyutiman!
Exactly, you need to roll it or you will be using future info.
actually would also be great to see an example application of this class. of course we can figure it out, but I think it'll enhance the article. just a suggestion.
Great article QB.
I have a question-- before orthogonalization, would you align all assets to the same time points by using each asset’s most recently completed bar, so every asset ends up with the same number of rows even though volume bars occur at different times?
Hey Jeff!
Not really, by doing that you would add lag. And due to the volume the length of each vector (per asset will probably be different as well)
suppose I have a futures series, instead of stock. the series is roll adjusted (by ratio), so the past prices are not historical prices, but they preserve the ratio p(t+1)/p(t). Would the fractional differencing work in that case?
Hey Dyutiman,
Fractional diff can work on a roll-adjusted futures series, but you should apply it to the log series, not the raw adjusted price level. It works only for research. For production it is a pitfall unless the series is constructed point-in-time