Also, some of the bumps that appear to be correlated are not bumps at all, but gaps in data – see 11/27 to 11/29. There are only a few data points during those dates, because my computer was shut down or something, and the script that grabs the numbers didn't run hourly like it usually does. So next time it runs, there's a big jump. I could do some math and divide the size of the spike by the duration of the gap, but I think I'll leave it as is.