Statistics are not just numbers: they require context to be useful
Statistical context includes definitions, methodological changes and other caveats that affect how we might otherwise interpret the figures.
Take crime statistics: if we look at the trend for the number of homicides recorded by the police in England and Wales there appears to be a large increase in 2002/03.
This was, in fact, the year that 172 historical murders carried out by Dr Harold Shipman were recorded by the police rather than a sudden increase in murderers on the loose; without these offences the trend could look very different. And without any supporting information, we wouldnt know that this was the reason behind the increase. The statistics could be misinterpreted.
There are numerous other examplesthe effect that the extra public holiday for the Royal wedding had on the economy or even how the birth of a celebrity child may impact the popularity of baby names.
Our early research suggests such caveats are currently handled inconsistently (and sometimes not at all) in the various formats that official statistics present themselves to us. Smart users are aware of this, but it can nevertheless destroy their confidence in using the statistics.
The work of former Full Fact secondees Louisa and Emily led to the publication of good practice guidance on presenting statistics in spreadsheets, which started to address some of the presentational inconsistencies of caveats in official statistics.
The next stage of work looks to expand on this to flag caveats in open data in general, to alert users to factors that help them to explain the data. This will include a review of key statistical releases to identify the types of caveats that accompany them, a typology of caveats being compiled and identifying best practice on how these should be coded and displayed in the data. Were looking to complete the work by the end of September 2015.
If youre interested in getting involved or for further information, get in touch with us at email@example.com.