Timeline for Finding causal relationship between two sets of time series data
Current License: CC BY-SA 4.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 29, 2023 at 0:15 | comment | added | Galen | "Causality in times series is a very challenging topic and is not as simple as applying some filter or adding lags to justify causal statements." Yes! (+1) And in what sense we think about causality matters. For example, from a potential outcomes framework or a Pearlian framework we would not think of Granger causality as necessarily being causal. | |
| May 17, 2023 at 20:26 | comment | added | Jonas Striaukas | I think first and foremost it starts with the hypothesis. Whether stock prices are caused by Google trends, in a structural sense, or not, well, not sure it would be meaningful to analyse. Granger causality would marely check for predictive relationship. | |
| May 10, 2023 at 14:33 | comment | added | Alexis | I think we agree that assumptions about causal structures are important (and Sugihara & Co. eschew this in their methods). Do you have a good recommendation for a nuanced philosophical take about Granger causality and prediction vs structural causal modeling? | |
| May 9, 2023 at 14:30 | comment | added | Jonas Striaukas | as in any causal analysis, it depends on the causal structure one is willing to assume. your last statement is, the very least, imprecise. Lastly, cross-mapping, like Granger, depends on the underlying causal mechanism. | |
| Apr 28, 2023 at 16:12 | comment | added | Alexis | Also Granger causality, is less about theories of causation, and more about prediction. | |
| Apr 28, 2023 at 16:10 | comment | added | Alexis | Sugihara and Co. disagree, and feel that empirical dynamic modeling, specifically convergent cross-mapping, does exactly what the OP asks for | |
| Mar 2, 2020 at 14:55 | history | answered | Jonas Striaukas | CC BY-SA 4.0 |