Crime

- For England, Wales and Northern Ireland we take 5 years of detailed crime statistics at individual crime level from https://data.police.uk/ (some 43 million crimes)
- We assign each of these the ONS LSOA (lower super output area)
- We then apply a rate to each of the LSOAs taking into account density of address points in the area
- We score each LSOA by binning the data into 50 buckets, we do this for each crime type which gives the _score variable
- We then apply an approximate normal distribution to classify the scores into 5 classes, 1 being the lowest, 5 being the highest, most data sits in 3 unless it is very low or very high, this gives the _class variable
- Finally we look at the historic data for each LSOA to see if there is evidence of a significant trend either increasing (1) or decreasing (-1), or no change (0), this gives the _trend variable
- For Scotland we do not have access to raw crime data, we use the data that they provide by SMID (Scottish Index of Multiple Deprivation) and apply a similar process as above to score and rank the data however we cannot split by class so all crime types will be assigned the same score nor do we have time series data available to calculate a trend so the _trend variable will be set to 0
- We take the LSOA + SMID and assign to postcode unit to create a full coverage layer across the UK

Note: some of our customers prefer to work with the simpler _class variable, others prefer the _score.  Just bear in mind that the distribution of classes will always work to an approximate normal distribution but the scores can change as the data changes in future

Note: postcode boundaries are provided by GeoPlan under copyright