# Asynchronous (irregular) Time Series Analysis

I am trying to analyze the lead-lag between time series of two stock prices. In regular time series analysis, we can do Cross Correlaton, VECM (Granger Causality). However how does one handle the same in irregularly spaced time series.

The hypothesis is that one of the instruments leads the other.

I have data for both symbols to the microseconds.

I have looked at RTAQ package and also tried applying VECM. RTAQ is more on a univariate time series while VECM is not significant on these timescales.

> dput(STOCKS[,]))
structure(c(29979, 29980, 29980, 29980, 29981, 29981, 29991,
29992, 29993, 29991, 29990, 29992), .Dim = c(6L, 2L), .Dimnames = list(NULL, c("Pair_Bid", "Calc_Bid" )), index = structure(c(1340686178.55163, 1340686181.40801, 1340686187.2642,
1340686187.52668, 1340686187.78777, 1340686189.36693), class = c("POSIXct", "POSIXt"), tzone = ""), class = "zoo")

• you need to use a reproducible set of data
– John
Commented Jun 26, 2012 at 9:53
• Not really sure why you say so? Can you elaborate? Commented Jun 26, 2012 at 9:56
• @John means (I think) that you are more likely to get a useful answer if you provide data that can easily be used by answerers to test and illustrate their methods (see tinyurl.com/reproducible-000 ). I would guess that parametric models for the cross-correlations/cross-spectra would be necessary ... Commented Jun 26, 2012 at 9:59
• this should really go on CrossValidated
– nico
Commented Jun 26, 2012 at 10:17
• because the question is probably sufficiently challenging that there isn't an obvious standard methodology. Rather than "I want to use well-known statistical procedure X, is it implemented in R/how do I go about using it?", this is more along the lines of "is there a good statistical procedure for solving problem Y"? Alternately, it might be worth checking out r-sig-finance (I think there is such a mailing list ...) Commented Jun 26, 2012 at 10:27