hamishgibbs / ghana_cdr_aggregation Goto Github PK

There are a few values with positive differences for journeys in the (all pairs network) - (sequential network). This should be impossible and likely results from fewer records in the sequential network (because journeys are less likely to rise above the censoring threshold as we hypothesis).

In that case - would it be better to use the sum, not the daily mean?

Here is an example:

Where value_....x is the all_pairs value and value..._.y is the sequential.

Clearly - what is happening is that this journey is only sampled for one day (presumably one day with high travel), then the sequential mean uses n=1 while the all pairs mean has a higher n.