We will look into the ZigZag approach for identifying waves and have identified 2 R packages that implement this approach:
- {TTR} package
- https://www.rdocumentation.org/packages/TTR/versions/0.24.0/topics/ZigZag
install.packages('TTR')
- {btutils} packages
- https://www.r-bloggers.com/a-better-zigzag/
install.packages('btutils', repos="http://R-Forge.R-project.org")
-
one thing that we will need to fix either conceptually or calculationally in the future: it is possible to end a wave without having a peak (because peak was < 10)
-
Can these notes be removed?
- need resurge annotation; e.g. CZE
- above annotation should persist through new peak; e.g. ISR
- not quite right post labelling e.g. LVA, PAK
- "peak" labelling wrong for PRT, (maybe SGE?)
- second wave resurgence in SVE and ISR
-
Errors
- None currently identified
-
Methodology notes
- We're not actually interested in valleys; on decline, we're interested in (1) point where first wave ends and (2) the trough (minimum value after the first peak but before the end of the second wave) - TODO PRIORITY
- Consolidation step - TODO
- SEE ALSO: summary of epidemiology of second waves here
- At the end of the time-series, classify dynamics for each country (typology) - second wave, resurgence, apparently controlled (MUS), other (how to describe LVA, PRT? simmering?)
- For second waves: what proportion of the time is the peak of the second wave higher than the peak of the first wave? (will be a minimum as all others are indeterminate until end of second wave)
- Time from first wave peak to end of first wave?
- Time from from end of first wave to trough?
- Time from first wave peak / end of first wave / trough to meeting second wave criterion?
- Time from first post-trough uptick to meeting the second wave criterion?
- How many upticks between end of first wave / trough and meeting the second wave criterion?
- How good an indicator of upswings / resurgences / second waves are upticks defined based on case numbers? positivity?
- What was the size of the first trough relative to the size of the first peak?
- How many total cases / deaths per million population by the end of the first wave?
- How many total tests per thousand population by the end of the first wave?
-
Figure notes (parameters: change = 15, minVal = 1, m = 14, len = 5, window = 8, threshold = 6)
- Visualization: second wave range should include 'above' and those below 1/3 of first peak but above 1/10 of second peak (until/unless resurgence) - TODO
- BEL: look into above definition - seems to occur at < 1/3 the first peak
- CZE: example of major and minor resurgences (if we want to define these)
- DEU: characterization probably the same as LVA (?), FIN - 'post-wave' (vigilance warranted)
- FIN: very jaggedy - seems to have a spurious upswing - is it also an uptick?; increase at the end would probably qualify as an 'uptick' in any non-technical assessment but doesn't meet our criteria (does mean upswing criteria at the moment; revisit this as definitions change)
- ISR: example of a resurgence during the second wave
- JPN: poor lead time from indicators
- KOR: example with upswings much earlier than second wave; should these be considered examples of false positive indicators? is there also a false negative? may need to define time windows - TODO
- LVA: very jaggedy - how would we characterize what's going on here? slow burn? post-wave?
- NGA: jaggedy without resurgences (but multiple non-wave peaks); probably indicative of poor surveillance, but how would we classify it - just as still in first wave (seems appropriate, actually)
- NLD: look into 'above' definition - seems to occur at < 1/3 the first peak (1/3 of second peak?); example of a (very minor) resurgence after which first wave ends - is this some form of false positive?
- PRT: jaggedy - has 3 peaks that aren't wave peaks and 2 short upswings and a resurgence
- SGP: example of a minor resurgence (2); example of upticks and upswing (no resurgence?) in positivity while cases going down
- SRB: when second wave drops below 1/3 of first wave 'above' definition no longer met but doesn't really have a meaning; change viz? (see also: AUS, others)
- SVN: look into 'above' definition - seems to occur at < 1/3 the first peak; example of major resurgence in second wave; example of a false negative (or is this just because of weirdness with 1/3 definition?)
- SWE: nice example of upticks indicating a resurgence (with current upswing definition, minor resurgence is dependent of including positivity)
- USA: nice example of upticks indicating a resurgence