1
$\begingroup$

In time series analysis we often deal with just one time series, so many time series analysis methods have been developed based on stationarity (see this question) .

However sometimes we can have sample of time series related to the same phenomenon.

To make an example, we could have many time series representing, each, the number of news media articles published about a particular event, say political elections, over one month. In this case we could have a data set comprising many time series of number of news media articles published on different elections disputed over time, maybe also across countries, and we could hypothesize that there is a common data generating process behind them.

Which kind of methods are available to analyze this kind of data set? Could you give me some hints about the keyword I have to search for to find something about this? If you want, I’d also appreciate some practical examples of approach to the analysis of data sets like the one in the example or a similar one (it is not a real case).

$\endgroup$

1 Answer 1

1
$\begingroup$

You can use Principal Component Analyses (PCA) to cluster multiple time series, see for example this questions. There are Python or R implementation available, for example, for Dynamic Factor Models (among others). The factors are time-varying and can be further investigated or used for forecasting. One drawback of this approach is of course that the factors have no direct economic meaning.

$\endgroup$
2
  • $\begingroup$ Interesting! I am also thinking how to exploit statistics with such a dataset. Having a sample of time series, one might calculate the sample distribution for each time point, as well as the mean and standard deviations (maybe by repeatedly resampling subsets of the original dataset and applying the Central Limit Theorem?). Thus one could calculate the sample(s) mean and sd considering them as those of the data generating process. In this way it might be possibile to get the “model” of the series, and calculate the series closer to the average and the extreme cases. What do you think? $\endgroup$
    – K9K9
    Mar 2, 2021 at 13:20
  • $\begingroup$ I realised that this is a different question, so I opened a new question here stats.stackexchange.com/questions/512095/… $\endgroup$
    – K9K9
    Mar 3, 2021 at 16:34

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.