Number of subscribers recorded in each region each hour/day, during weekdays: this describes how busy each region is during weekdays, and during each hour of weekdays

Number of subscribers recorded in each region each hour/day, during weekends: this describes how busy each region is during weekends, and during each hour of weekends.

The variance in hourly/daily patterns, across different days/weeks: this provides information about the extent to which previously predictable, regular patterns have been disrupted.

Disruption duration: A further indicator may be the time it takes for presence patterns to become regular again.

Ratio of the number of subscribers recorded in each region during ‘daytime’ hours, and the number during ‘nighttime’ hours: this provides information about the number of residents (who are likely to be present during the nighttime) versus the number of non-residents (who are likely to be present during the daytime) that visit the region. It can also provide information about which regions are popular night-life spots (a high proportion of nighttime visitors).

Number of subscribers that are assigned to a region as their home region.

Number of subscribers whose home location has changed, between any two regions or for specific regions.

Average time for a region to get back to its pre-crisis subscriber population size (if at all).

Number of subscribers recorded at each cluster within a 15 minute interval (urban areas only): this measures the size of a ‘crowd’.

Number of ‘crowds’ (defined to be more than X subscribers recorded at the same cluster within a 15 minute interval) recorded each day: this measures the frequency with which crowds congregate.

Number of unique subscribers recorded in the region each day/week, normalised by the number of subscribers that have that region assigned to them as their ‘home region’: this is an indicator of how many non-resident ‘outside’ visitors come to the region, relative to the size of the resident population.

Average number of unique visitors each day, averaged over a week, divided by the number of unique subscribers to the region in a week: this indicates the degree to which the same people visit a region each day, versus different people visiting each day which would result in higher ‘mixing’.

Average number of sub-regions visited in each region per visitors to the region: this measures the amount of movement within a region.

Number of subscribers entering a region from another region.

Number of subscribers leaving a region and travelling to another region.

The difference between incoming and outgoing flows: this measures how subscribers are redistributing amongst regions.

The median distance (and quartiles) travelled by all subscribers that travel to the region from another region: this indicates whether a lot of people come to the region from nearby regions, or from regions that are farther away.

Can be aggregated across regions.

For each destination region, this is the average percentage of subscribers that originated from each region: this indicates whether a lot of people originate from the same region, or whether there are many origin regions with a few people coming from each one.

Variant: standard deviation of coordinates of origin regions.

Average ‘mixing factor’ of origin regions: indicates whether visitors are coming from a ‘high-risk’ (or ‘high-contribution’) region.

Average number of regions visited by the resident subscribers of each region.

Variant

Exposure (or contribution): changes to the number of hotspots visited, per home location

Average number of hotspots visited by the resident subscribers of each region)

Average distance to visited regions by the resident subscribers of each region.

Standard deviation of coordinates of visited regions by the resident subscribers of each region.

More specific indicator on the number of subscribers not crossing admin boundaries.

More specific indicator on the number of subscribers not recorded outside of their home region.

Are there areas of the country isolated for the rest? (which may be easier to protect).

List of locations most connected to each other forming a path through the country (along which the virus may spread).

Path along which most population mixing is occuring (along which the virus may spread).

Locations with high population mixing values and/or crowdedness.

The total number of calls/SMS/data sessions in each region each hour/day.

This is to provide an indicator of the sample (data) size that we are working with in order to know if the analysis for particular regions and times should be discounted due to insufficient data. It also enables to check whether an increase in ‘subscriber counts’ may actually just be due to the normal number of subscribers using their phone more, and to easily determine if there is an issue with e.g. missing data one day.

The proportion of residents in a region who use their phone each hour/day, divided by the total number of residents (where ‘residency’ is defined via a home region): this indicates how active the subscribers in each region are.

Percentage of all subscribers who are ‘new’ (appeared in the dataset for the first time within the last week): this provides information about how much data we have for subscribers (to inform statistical validity of analyses).