Honolulu and Maui's weekly trends, from 8/16 Sunday to 8/22 Saturday, show a unique, linear pattern that rarely happened in the past month (except Maui CCC on 8/19 Wednesday). As a result, the semi-log linear regression is below the HICOVID model.
If we follow the 7-day linear regression trend shown below, one can expect that Honolulu CCC would reach a little over 6,000 by 8/23/2020. In this stage, all the extrapolating models of linear and semi-log linear regression, and the HICOVID models show similar trends within a narrow range. The hyperbolic tangent model still significantly underestimates the forecasting values in comparison to the other three models. The Honolulu CCC is still in a growing phase, influenced by the previous causes, which have not been fully identified.
What about next week and next month? The weekly increaseof Honolulu CCC show how many new cases are reported this week (0), two weeks ago (-2), and so forth. Honolulu has 1,416 new cases in this week from 8/16 Sunday to 8/22 Saturday (today). In the last week, the number is 1,426. In the semi-log plot (in red), the difference of 10 is not noticeable; but the linear plot (blue) shows a small decrease. It would be terrific if the Honolulu CCC would decrease in the opposite manner of its growth in the past weeks: for example, next week, it drops to the level of 8/8. Comparing this week and the last week, we have ten less in new cases: and comparing the last week and the week before the last week, we had 1, 426-1,122 = 304 more cases. Therefore, one can interpret that the Honolulu CCC starts increasing slowly by having lesser daily new cases in the following weeks.
On the other hand, the semi-log plot (in red, on the left-hand side) gives us a possibly opposite conclusion. There were two instances when the weekly new case number decreased, i.e., weeks of indices -8 and -5, indicating 8 and 5 weeks ago from now. A week later, in each case, the Honolulu CCC increase much more than the previous weeks. There is a high probability that the Honolulu CCC will rise again, which we hope not to. It is tough to forecast next week's increase in the Honolulu new cases. It is also possible that the re-rising would not be so severe, because the weekly data of this and the last week are very similar.
It is too early to conclude that the Honolulu situation becomes better by analyzing the two-week data until now. If you look at the red semi-log plot, it does not show that the Honolulu CCC has stopped increasing exponentially in the last ten weeks. UH instruction will be almost 100% online, which fundamentally prevents social interaction. K-12 schools should also pay close attention if any in-person instructions are planned. The recent restriction announcement is very timely, but at the individual level, staying home as much as possible without interacting with anybody except people living together is the best and only option, and it should be systematically done without too many exceptions.
The CCC data for the Honolulu county seems to be following the 7-day semi-log linear regression rather than the prediction curve 1st, 2nd, and 3rd curve as if it was following the orange curve the daily CCC would be starting to plateau today.
The slope of Honolulu's weekly trend graphs linearly increases with a constant slope except for a minor dip in the line on Tuesday.
While the number of cases between Saturdays is staying constant for Honolulu (about 1,400 cases), the number of cases between weeks on Saturdays for Maui has significantly been fluctuating. From 8/15 - 8/22, there was an increase of about 70 cases. We observe that Maui also is entering a new spreading phase. More analysis comes.