Discussion on Papers 4 and 5

Mr Penrice (Paper 5)

The reliability of data must always be examined before any attempt is made to base conclusions on them. Manual enumeration has sizeable observer-variability in daily traffic counts, but the errors seem less horrific after allowing for the number of days covered and the variability of traffic flows; flows at individual points often increase seasonally by three to four-fold.

Vetting the accuracy of data is often unduly neglected, but there is the danger of over-vetting. Vetting procedures that were sensible when introduced have sometimes been carried on mechanically, resulting in series which were worse than those unvetted. There is also the danger of being too over-sold on linked data; their errors often accumulate rather than cancel out.