In this introductory lecture, research fellow Barry Rollingson explains the role of spatial statistics in health informatics, tracing its roots to John Snow’s historic cholera map as a classic example of a spatial point pattern. He outlines three major data forms used in spatial analysis: point patterns, areal (administrative unit) data, and geostatistical samples, describing their characteristics, typical applications (epidemiology, pollution monitoring, mining, cellular and cosmic studies), and the analytical questions they can address. The session concludes by previewing the first practical exercise, which will guide learners in generating and visualizing random point patterns in R.
Chapters
Hello, I'm Barry Rollingson and I'm a research fellow in the Centre for Health Informatics, Computing and Statistics Che CAS at Lancaster University's medical school. We use spatial statistics to look at health data, to detect epidemics, to advise on treatments and to spot risk factors. Our work harks back to the early days of epidemiology, when John Snow, the Victorian doctor and not the character from a popular fantasy series, produced this map of cholera deaths in London by marking each house with a number of cases, like a bar chart. This showed an association with a water pump, giving weight to his theory that it was a waterborne disease. And if you're ever in London, you can celebrate his discovery in the pub that bears his name on the corner where the pump was. Cheers. The cholera cases are an example of a spatial point pattern. These events have happened in these locations and the question is why they happened there and not elsewhere? Is there some association with the environment? Are the cases spreading? Are some people more susceptible? Given the right data, these are the questions that point pattern analysis can answer. Point pattern analysis Whether locations are of statistical interest is just one aspect of spatial statistics. Spatial data comes in other forms too. Sometimes data is made available for each of a number of divisions of a larger area. Most population statistics from government or health agencies are presented for states, countries or other administrative units. These units can be odd shapes, varying sizes, different demographics. And conclusions from data like this need to consider how these areas relate to one another. If what you have is a set of samples taken at a set of survey point locations, then you have geostatistical data. The locations themselves aren't of interest. Presumably you chose them for some good reason, but the measured values are. Perhaps you have pollution samples from testing stations and you want to know about airborne particulates at schools. Or you're taking soil samples and you want to know where the gold is. Geostatistical theory came mostly from mining and oil exploration. But spatial statistics doesn't have to be limited to data or on or in the Earth. It can be used to study the microscopic too, such as the pattern of cell formations in tissues or the cosmic to test hypotheses about the distribution of galaxies in the universe. But let's keep our feet on the ground. For the first set of exercises, you'll learn how to create some random point patterns and plot them.
