|
|
|
|
|
|
by iflint
1409 days ago
|
|
|
Be careful using the YoY number, you need to factor in the base effect. (i.e. the change in YoY number each month factors in the new month of data but also the month 13 months ago that is no longer included) MoM is a cleaner way to look at things in volatile times like these. |
|
|
I realized when monitoring COVID data that rolling numbers like this can be heavily impacted by the number that’s rolling off.
Take, for example, a 7-day COVID case count of 100. The next day this drops to 90. That’s great, right?
Well, if yesterday’s 7-day values were all 100, and today’s are 100, 100, 100, 100, 100, 100, 30, then yes, that’s great!
But if yesterday’s 7-day values were 700, 0, 0, 0, 0, 0, 0, and today’s are 0, 0, 0, 0, 0, 0, 630, then that’s potentially very bad (depending on other context).
This is obviously a hyperbole, but it shows that looking at and comparing rolling numbers can be tricky, and you need to look at other context to get the full picture.