{"id":2949,"date":"2021-05-03T14:37:55","date_gmt":"2021-05-03T02:37:55","guid":{"rendered":"http:\/\/blogs.lincoln.ac.nz\/gis\/?p=2949"},"modified":"2023-05-07T03:05:44","modified_gmt":"2023-05-07T03:05:44","slug":"deleting-smokers-closing-the-loops","status":"publish","type":"post","link":"https:\/\/blogs.lincoln.ac.nz\/gis\/deleting-smokers-closing-the-loops\/","title":{"rendered":"Deleting Smokers – Closing the Loop(s)"},"content":{"rendered":"

The last of our series on developing a Python script for some smoking reduction analysis.<\/em><\/p>\n

Long term readers of the GIS Blog (Hi Mum!) may recall a series of posts on the development of a Python script for some analysis.\u00a0 As part of the smoking reduction analysis I’ve been involved with over the years, we wanted to look at the impacts on accessibility to tobacco from imposing some distance limits on tobacco retailers.\u00a0 In other words, how would setting a minimum distance between retailers affect how easy it is to buy tobacco?\u00a0 After a survey of the literature, we settled on three distances to evaluate: 450 m, 300 m and 150 m.\u00a0 The previous posts set up the background<\/a>, tweaking some psuedo-code<\/a> and developing one of the key data preparation<\/a> scripts.\u00a0 In this post, we put this script to bed (or rather, get it out of bed and running!) and get some results.<\/p>\n

\"\"<\/a><\/div>\n

So as a quick recap, we needed a script that would look at the 5,131 existing tobacco retailer points we’ve got in our existing database, figure out out close they all are to each other and then sequentially go through each, defining which ones are within a given minimum distance and then randomly delete one of them, and then keep doing this until none of the retailers is within that minimum distance.\u00a0 And then<\/em>, do it all over again 1,000 times so we end up with a 1,000 iterations of this – giving us an ensemble of possible configurations.\u00a0 The random deletion means that everyone could be different and we have a massive dataset to do some more general statistics on.<\/p>\n

Well I’m very pleased to say that we’ve brought this all to fruition, though with 100 iterations rather than 1,000.\u00a0 In this post I’ll try to highlight the key parts of this script without getting too bogged down in the details (Ed.: a tall order<\/em>).<\/p>\n

In one of those earlier posts we saw the first data preparation script to find the proximity relationships between pairs of points.\u00a0 With those, it’s then a matter of going through each pair and randomly deleting one of the two.\u00a0 The tricky bit is going through the whole data set once, randomly deleting a point in a pair, but then doing it over and over again until none remain within that set distance.\u00a0 This immediately suggests loops that continue until a certain condition is met.\u00a0 The Near tool has been a critical tool in this script as it not only allows us to calculate the straight-line distance to the nearest retailer, but also allows us to limit how far away to look.\u00a0 So if our minimum distance is 450 m, any retailer with no other retailer within 450 m will be assigned a -1.\u00a0 When all<\/strong> the retailers get a NEARDIST value of -1, then that iteration is done and the script can start all over on a new sequence.\u00a0 So overall, we’ve got three main loops to this script:<\/p>\n