Research and Analytics

Nearest Neighbours Model: Methodology Note and Instructions

Nearest Neighbours Model: Methodology Note and Instructions

Posted on 15th November 2005; updated on 11th July 2008

The Approach

Comparative analyses between 'subjects' can be drawn by means of a number of data reduction techniques.

Nearest neighbours is one of them and it follows the traditional 'distance' approach. Each of the variables is standardised (with a mean value of zero and standard deviation of one) and Euclidean 'distances' between all possible pairs of local authorities is calculated.

To arrive at the final distance measure between authorities, these distances are then summed across for every single indicator and 'rebased' by assigning a distance of 1 to the farthest neighbour. Therefore, all overall distances will then lie between zero and one.

Note that the model also allows you to select the number of 'nearest neighbours' to be shown, so you can display the top 15, top 30 or all neighbours.

The output returned by these calculations is a simplistic way of presenting a fairly complex underlying idea. Broadly speaking results and common sense go hand in hand reasonably well. Nevertheless, the outcome relies on the indicators and mathematical procedures used.

Running the Model

1. To run the model, firstly select your authority using the drop-down menu;

2. The model will identify your class of authority (e.g. Inner London) and will automatically include this class as a 'selected comparator'. To add further classes, simply click on the appropriate class and click on the single arrowed button. The class title should switch over to join the existing comparator group;

3. At the next stage you will see a full list of the variables available to calculate your neighbours group. Some default variables are ticked indicating that they are utilised. If you wish to customise this list you may remove this tick (by clicking on it) for any of the variables to exclude them from the calculations or, alternatively, you may want to select some other unticked indicators;

4. Finally you will be asked to select the size of your group.

Hitting the button marked 'submit' will run the model and generate the list of neighbours. The top authority represents your nearest statistical neighbour (shown numerically and represented graphically by the size of the horizontal bar).

Related files:

Database File Download the nearest neighbours methodology

Related links:

Public Libraries
Trading Standards

CIPFAstats Current Datasets