I'm not that familiar with time-series type data. I am looking for some advice on the interpretation of the following plots of autocorrelation between two variables. I used ccf in R to assess how two variables 'x' and 'y' are associated with each other over time. I did this for several separate datasets of 'x' and 'y'.

Each dataset has nearly 2000 observations, although there are some missing data points (usually at the beginning of the dataset, but not too numerous).

Example code would be:

ccf(x, y, na.action=na.pass)

Here are some example plots: enter image description here

My questions are:

1) Interpretation. For plots 1 and 3 there are lags where the correlation is significant at both a negative and positive lag. What is the best interpretation here? Should one simply focus on the largest correlations? Are there better methods of understanding which direction 'x' and 'y' are associated?

2) The effect of missing data. Given that only approximately 30-40 values of either 'x' or 'y' are missing on average from the beginning of a 2000 observation dataset, I'm assuming that this does not overly affect results - but is this accurate.

I hope these questions are clear. Again, this isn't my usual form of data analysis, but I'm interested in any comments.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.