Spain and Italy time series correlation

Abstract

Different countries attacked the COVID-19 outbreak in different ways; however, the results seem to have some correlation, even more so if the countries are similar geographically, politically and culturally. In this report we investigate correlations between the time series of infection and deceases in Spain and Italy, and try to come up with a lag between them, so that we can infer something on the Spanish situation, which seems to go behind.

Introduction

Information on possible future evolution of the pandemic gives the citizen a bit of control over what’s going to happen to their lives (Liu et al. 2020). Since different countries seem to be in different stages of the epidemic, and Italy is substantially similar to Spain, we will try to see if those similarities translate in correlation between the different time series, and, in that case, if the lag is similar to the perceived lag of one week.

Absolute correlation will be investigated first.

The cross-correlation is evident, and substantially similar in profile, and but it goes from -11 days (Italy leading Spain) to +8 days (Italy lagging Spain). This is probably due alto to auto-correlation of the time series, which provokes this effect.

New cases are probably a bit more revelatory:

Again, same effect, only extended in time and skewed to Italy leading Spain for up to 12 to 14 days. The peak is at 0, anyway, so the best predictor of one time series is the other time series the same day.

Conclusions

Correlation of time series in Spain and Italy do not yield a clear result. Although it is significant, and the distribution of cross-correlation values is slightly skewed to the left indicating Italy leading Spain, it can’t clearly be said that there’s a difference of one week between them. If we make the difference between Italy leading Spain and the inverse, it’s around 6 days for the deltas, 5 days for the absolute time series; that is the only hint of this difference, but it would be cut short by one day to the guesstimate of 7 days.

At least, we have a statistical confirmation of this relation, and by checking the situation in Italy, we can at least reduce uncertainty of what the situation in Spain is going to be in approximately 6 days.

Acknowledgements

This file has been generated from data published by JHU CSSE. It’s data-driven and it can be re-generated from the script in this repository.

License

This report has a cc-by-sa license, and its code a GPL license. Reviews, suggestions and patches are welcome. It’s written in RMarkdown and hosted in GitHub.

References

Liu, Kai, Ying Chen, Duozhi Wu, Ruzheng Lin, Zaisheng Wang, and Liqing Pan. 2020. “Effects of Progressive Muscle Relaxation on Anxiety and Sleep Quality in Patients with Covid-19.” Complementary Therapies in Clinical Practice. Elsevier, 101132.