5 ms·
For the data scientists- what would it take to prove that there is/isn't a statistically significant correlation (not causation) between Mercury retrograde and
by uptownfunk 5y ago
For the data scientists- what would it take to prove that there is/isn't a statistically significant correlation (not causation) between Mercury retrograde and tech large scale events / SEV1's
- twofornone 5y agoDepends on how many cases of "large scale events" you have. You could calculate a simple correlation coefficient and even a p-value to loosely estimate causation, if you had enough data points.
- barbazoo 5y agoAs a noob in statistical analysis I'm wondering how would one calculate the correlation coefficient between two sets of dates (occurrences of something happening, in this example outages/SEV1s and days of Mercury retrograde)?
- RONROC 5y agoI guess the grand irony of all this is that data scientists — of which, the "good" ones (rational Spock-types, highly technical, sometimes to their own detriment) are, historically, not socially inclined — are the ones responsible for optimizing _social_ media platforms in the first place. But yeah, something something Mercury.
- uptownfunk 5y agoI think what you could do (scientists please feel free to correct me) is you basically separate days into two groups, 1/ days falling during mercury retro 2/ days falling during not-mercury retro (I think called direct motion). Then you look at the total LSEs/total days for each of the two groups and do a two-sample t-test...? Depending on the p-value you can conclude there is some correlation. Now for the fun part.. data acquisition. Anyone know a good source for tabulating LSEs/day?