More on the turn in proportions testing positive in Gauteng (mentioned in last night's thread:
1. While we had suspected that the daily PTP had maxed out on 1 Dec, the clear periodicity in the data means that we had to wait to confirm
https://twitter.com/tomtom_m/status/1471137248259317762).
1. While we had suspected that the daily PTP had maxed out on 1 Dec, the clear periodicity in the data means that we had to wait to confirm
1b. We also had crashed into data problems of loaded data from public facilities, so had to split by public/private to be more certain that PTP in each had turned before calling it.
On top of that, the known issues with loading of Ag test results means focus must be on PCR tests.
On top of that, the known issues with loading of Ag test results means focus must be on PCR tests.
1c. Weekly periodicity means that it's preferable to use a 7day centred moving average, and to wait until the daily peaks had moved out of the 7DCMA window to confirm. Adding another day's data (covering to 14/12) confirms the trend. NB! GAUTENG PCR ONLY 
1d. Data for 15/12 not yet incorporated as specimens collected yesterday still being loaded, while we know that most PCR test results are loaded within 24 hours.
2. As noted previously, it is also essential to look at data by collection date (and preferably PTP by collection date, to get rid of volume effects), and not by the date of reporting. This graph shows why:
NB GAUTENG ONLY
NB GAUTENG ONLY
NB GAUTENG ONLY
RHscale
Black line: 7-day centred moving average of PCR PTP.
Grey line: daily PCR PTP (by date of collection)
LHscale
Blue line: daily PCR cases (date of collection)
Red dotted line: daily ALL cases (as reported nightly)
Red line: 7-day CMA ALL cases (reported)
RHscale
Black line: 7-day centred moving average of PCR PTP.
Grey line: daily PCR PTP (by date of collection)
LHscale
Blue line: daily PCR cases (date of collection)
Red dotted line: daily ALL cases (as reported nightly)
Red line: 7-day CMA ALL cases (reported)
NB GAUTENG ONLY, and the *reported* (red) data includes Antigen AND PCR tests. The other data are PCT only
The red line (7DCMA ALL cases, reported) peaks on 11/12, a full week after PTP (7DCMA, collected) peaked on 4/12. Daily PTP (collected) peaked on 1 December ...
The red line (7DCMA ALL cases, reported) peaks on 11/12, a full week after PTP (7DCMA, collected) peaked on 4/12. Daily PTP (collected) peaked on 1 December ...
NB GAUTENG ONLY
the reported (both PCR and Ag) cases peaked only (possibly!) on 12 December (partly due to to the IT problems last week and delays in loading Ag cases)
the reported (both PCR and Ag) cases peaked only (possibly!) on 12 December (partly due to to the IT problems last week and delays in loading Ag cases)
Point being: Followers of South African Covid data should be careful of reading too much into analyses based solely on the nightly reported data. Yes, the patterns are broadly similar; the subtle differences can be marked.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
