The other day I was catching a bus home later at night, which made me acutely aware that I should not take the frequent daytime transit in Vancouver for granted. On the ride home I decided to dig into this and grab some transit data. We have played with transit data before, but since this was going to be the second time it was high time for a quick R package to standardize our efforts and simplify things for the next time around.
NHS Income Data, a First Retrospective There was much hand wringing when NHS income data got released. The change in methods were big, most notably the replacement of the mandatory long form census, that was administered to a random 1 in 5 sub sample, by the voluntary NHS that went out to approximately 1 in 3 households. The (design-weighted) response rate for the NHS was 77%, compared to 94% for the long form in 2006.
Vancouver’s median household income has grown. But there are many ways how this could have happened. We want to take a deeper look to understand how the income distribution changed. To that end, we will investigate the change in the number of people in each income bracket between the census years. And put that into context to what happened in the region and Canada wide. This is a mixture of what we have done when comparing the size of age groups between censuses.
I started writing this blog post in December 2015, when CensusMapper quite a bit younger and I hacked together some basic dot-density maps. I never much liked the results and have been slowly improving and thinking about them. I am still not entirely happy with the current implementation, but it is slowly getting there. The final impulse to finsish this post was the work on cancensus, and R wrapper for the CensusMapper API my explorations in multi-category dot density maps in R, now tied up into the new dotdensity package.
At CensusMapper we like building models based on census data. We now have a common tiling for 2011 and 2016 geographies that allows us to easily model changes over time. After building a model we often want to see how well the model performs. An easy way to do this is to simply map the difference of observations and model predictions. Those maps are great and it is easy to understand what is mapped.