3 ms·
And it's not even in the article title. In fact, the text says " Figure 1A compares the annual incidence of the disease and the household equipment rate in sev
by uniqueuid 5y ago
And it's not even in the article title.
In fact, the text says " Figure 1A compares the annual incidence of the disease and the household equipment rate in several countries. There is a temporal correlation between these variables."
A temporal correlation is something entirely different to a cross-sectional correlation (arrow of time etc.)
- DantesKite 5y agoI tried researching what a temporal correlation is. Do you have short explanation for it?
- uniqueuid 5y agoTemporal correlations are a family of things, since there are many degrees of freedom to choose what to compare with what. A very simple test is the "Granger Causality", which tests whether one timeline significantly predicts another one. You can formulate more complex models such as time series a predicting b with a certain lag. Ultimately, the idea is most often to remove unrelated factors (such as control variables, seasonality, self-influence i.e. autoregression) and then measure how well one series at t(0) predicts another one at t(1), while optionally doing some sort of hyperparameter optimization for the lag (i.e. determine which lag works best).
- DantesKite 5y agoThank you. Much appreciated.
- jldugger 5y agoA temporal correlation is basically saying that the correlation is strong when broken down by time. Of course, causation is trickier; roosters don't cause the sun to rise, even if they occur simultaneously (or even slightly before!).