Suppose I'm measuring concentrations of gas over time as well as relative humidity. How can I find the correlation between time series? The author suggests Pearson's correlation and the Cross Correlation Function and has a Ph.D., so I assume he knows what he's talking about. Is the CCF in R okay to use?
Some of his quotes are:
"The Pearson correlation measures how two continuous signals co-vary over time and indicate the linear relationship as a number between -1 (negatively correlated) to 0 (not correlated) to 1 (perfectly correlated). It is intuitive, easy to understand, and easy to interpret. Two things to be cautious of when using Pearson correlation are that 1) outliers can skew the results of the correlation estimation, and 2) it assumes the data are homoscedastic such that the variance of your data is homogenous across the data range. Generally, the correlation is a snapshot measure of global synchrony. Therefore it does not provide information about directionality between the two signals, such as which signal leads and which follows."
"Once again, the Overall Pearson r is a measure of global synchrony that reduces the relationship between two signals to a single value. Nonetheless, there is a way to look at moment-to-moment local synchrony using Pearson correlation. One way to compute this is by measuring the Pearson correlation in a small portion of the signal and repeating the process along a rolling window until the entire movement is covered.
"Overall, the Pearson correlation is a good place to start as it provides a straightforward way to compute global and local synchrony. However, this still does not provide insights into signal dynamics such as which signal occurs first, which can be measured via cross-correlations."
"Time-lagged cross-correlations and windowed time-lagged cross-correlations are a great way to visualize the fine-grained dynamic interaction between signals such as the leader-follower relationship and how they shift over time. However, these signals have been computed assuming that events are happening simultaneously and in similar lengths."
Do you think all four methods in this article are okay to use? The comments suggest they do, but I want to check opinions here.
https://towardsdatascience.com/four-ways-to-quantify-synchrony-between-time-series-data-b99136c4a9c9
