The following block of code maps these counts by area. Both equal interval and quantile classification schemes of the same data are mapped. However, as we shall see in the following sections, this perception may not reflect reality. We therefore seek to produce perceptually tenable maps.
|Country:||Bosnia & Herzegovina|
|Published (Last):||14 June 2008|
|PDF File Size:||10.17 Mb|
|ePub File Size:||12.36 Mb|
|Price:||Free* [*Free Regsitration Required]|
The following block of code maps these counts by area. Both equal interval and quantile classification schemes of the same data are mapped. However, as we shall see in the following sections, this perception may not reflect reality. We therefore seek to produce perceptually tenable maps. Dykes and Unwin Dykes and Unwin define a similar concept called map stability which seeks to produce maps that convey real effects.
Raw Rates A popular approach for correcting for biased visual weights due, for instance, to different unit area sizes is to normalize the count data by area thus giving a count per unit area. Though this may make sense for population count data, it does not make a whole lot sense when applied to mortality counts; we are usually interested in the number of deaths per population count and not in the number of deaths per unit area.
In the next chunk of code we extract population count under the age of 5 from the Auckland data set and assign this value to the variable pop. Likewise, we extract the under 5 mortality count and assign this value to the variable mor. Bear in mind that the mortality count spans a 9 year period. Since mortality rates are usually presented in rates per year, we need to multiply the population value which is for the year by nine.
This will be important in the subsequent code when we compute mortality rates. Both quantile and equal interval classification schemes of the same data are mapped. Note how our perception of the distribution of infant deaths changes when looking at mapped raw rates vs.
Standardized mortality ratios relative risk. Another way to re-express the data is to map the Standardized Mortality Ratios SMR -a very popular form of representation in the field of epidemiology. Such maps map the ratios of the number of deaths to an expected death count.
There are many ways to define an expected death count, many of which can be externally specified. We can therefore apply a diverging color scheme where green hues represent less than expected rates and red hues represent greater than expected rates.
This next chunk of code maps the under 5 population count by census area unit. Interestingly, the three highest raw rates in Auckland One approach to circumventing this issue is to generate a probability map of the data. The next section highlights such an example. The a priori estimate can be based on some global mean. An example of the use on a global EB infant mortality rate map is shown below.
The EB map is shown side-by-side with the raw rates map for comparison. Unstable rates i. Notice how the three high raw rates highlighted in the last section are reduced from The adjusted estimated rates can be shrunk towards a local mean instead. Such technique is referred to as local empirical Bayes rate estimates. In the following example, we define local as consisting of all first order adjacent census unit areas. References Bailey, Trevor C. Interactive Spatial Data Analysis.
England: Prentice Hall. Dykes, J.
Mapping rates in R
Interactive spatial data analysis