Open rcyeh opened 11 years ago
Suppose I want to ask the following:
What I want to do is: for any stock, calculate the value of the VPIN indicator at various instants, and correlate the value of the indicator with forward returns over some horizon. For the purposes of this exercise, let's first choose 1-minute bucketing and 1-minute forward returns; and then later we can go back and try volume bucketing or other horizons or both.
What I need to do is: for each trade, estimate whether it was a buy or a sell order. I do this by comparing the trade price with a recent mid-market price. Once all trades are labeled, then aggregate them in time (or volume) buckets.
Does this make sense?
Let's practice on a small amount of data to get started. I just uploaded a new file to s3://cfem2013/ticks.20130423.h5, which has ticks organized by symbol. I've confirmed that this file can be read in python (pytables) and R (rhdf5), and will add a guide as to the information in this file.
Hi Richard, thanks a lot for these inputs. We understand what need to be accomplished. I uploaded a very brief outline, which is pretty much the same as your initial proposal, plus some thoughts. I will try to start parsing the data this weekend. May I ask what period does the data cover? (Perhaps it will be apparent once I start the parsing) For everyone else, the file is on our amazon virtual machine. Feel free to do anything with it.
As you may have noticed, we are not very responsive recently, as many of us are on the last week of our internship and are wrapping up intern projects, and some of us are also trying to find short term lease for the next month, trying to renew their student visas...etc.
I apologize for the inactivity. As we are finishing up our internship, we will become more active in advancing the progress.
The data are only for a single day: April 23, 2013. Filed #4 and #5 as necessary tasks for the next step.
Please expand your outline into github issues for each significant step in the process. That is, make one github issue per remaining step, as I did with documenting the tick file and installing software.
OK, I see your recent commits. I think it's a step in the right direction.
In reference_data/construct_vpin_and_analysis.txt, you wrote that you need transaction-level data for a liquid index product. Next time, please raise that as an issue on github. How about symbol SPY?
SPY is fine. Can we also get tick data for VIX options?
(Opened #7)
We begin with #2 (Summarize VPIN).