Issue 5 The statistical analysis
Why would you attempt to reconcile two different sets of opinions, one based on the views of people who live in Riseley and have nothing but good intentions for the village and the views of a planning consultancy, AECOM, who have never spent a full 24 hours in Riseley and who have no long term commitment to Riseley and its future. The way to reconcile the different views is not to run a complex statistical analysis but to go back to the people of Riseley, in a proper public consultation and ask them what they think, what would they prefer.
I find it surprising and inappropriate that statistical analysis has found its way into the Riseley Neighbourhood Planning process and I disagree with the conclusions drawn from the analysis. It would be helpful to know how many other Neighbourhood Plans have used Spearman's rank correlation.
There may be a statistical correlation between the Riseley Neighbourhood Planning Group rankings and the AECOM rankings, of site suitability. However, the Spearman's rank correlation of 0.71, 0.73 or 0.67 depending on which data set is used, does not mean that the interpretation is correct for all sites . Some sites, in statistical terms are "outliers". They don't fit the pattern and they should be evaluated differently.
The statistical analysis appears objective but is actually very subjective.
Based on the statistical analysis, it would seem that the Riseley Neighbourhood Planning Group are asking us to accept that a Riseley Neighbourhood Planning Group ranking of, say 7, means the same as an AECOM ranking of 7. Given the long list of differing assessment criteria used by the Riseley Neighbourhood Planning Group and AECOM, this cannot be considered as an assumption that has been met. In other words, correlation and agreement analyses in this context, are for measuring the same quantity with two different ranking systems…is rank=7 the same ‘quantity’ with either ranking system? Therefore, the measuring of correlation may not work as intended, and particularly, the existence of outliers, ranking differences can probably be explained more by the different ranking criteria used by Riseley and AECOM, rather than the person/s making the ranking? The people of Riseley should be given the chance to properly evaluate and rank Site 512, The Paddock.
There is a common statistical convention to determine an outlier. Values beyond 1.5 times the interquartile range, IQR, are usually considered outliers. Using this convention and looking at Table 8 in the Analysis of Site Suitability assessments conducted by Riseley and AECOM, four sites can be considered as outliers, Site 211, 20 Rotten Row, Site 219, Riseley Lodge Farm, Site 614 Land at Town Farm, Lowsdon Lane and Site 512, The Paddock. It should be noted that Sites, 211, 219 and 614 have all been listed as potentially suitable for inclusion in the Neighbourhood Plan, subject to the mitigation of identified constraints and/or consultation with Bedford Borough Council, only Site 512, The Paddock has not. So of the four sites that don't quite fit the pattern in Table 8, three are potentially allocated and one is not. If you look at Tables 9 and 10 , Site 512 is always is an outlier, not fitting the pattern. This is reason to revaluate Site 512, The Paddock. Riseley Parish Council shouldn’t judge Site 512 just by the overall correlation between the local Riseley assessment, where it came 4th and the AECOM assessment where it came 12th (Table 8). There should be a specific decision-making scheme to take into account other factors and Site 512 should be allocated for development in the Riseley Neighbourhood Plan subject to consultation with Bedford Borough Council.
Analysis of Site Suitability assessments conducted by Riseley and AECOM,
The outliers are the sites where column ‘d’ (=difference between AECOM and Riseley rank) has a value of greater than IQR X 1.5
Table 8
Sample size:19
Lower quartile (xL): 1
Upper quartile (xU): 4
Interquartile range (xU-xL): 3
3 X1.5 = 4.5
Outlier sites: 211, 219, 614 and 512
Table 9
Sample size:19
Lower quartile (xL): 1
Upper quartile (xU): 5
Interquartile range (xU-xL): 4
4 X 1.5 =6
Outlier sites: 211,219,218 and 512
Table 10
Sample size:19
Lower quartile (xL): 1
Upper quartile (xU): 5
Interquartile range (xU-xL): 4
4 X 1.5 = 6
Outlier sites: 211, 219 and 512