could also explore this option from

@JonnoB

https://github.com/JonnoB/BigHeat/blob/master/CompressDF.R

To visualise the missingness, I made a heat map ordered by hierarchical similarity. However, when I visualised it I had two problems. The first was that the size of the dataframe was so large that ggplot started choking up. The second problem was that when visualised the heatmap could be difficult to understand as individual points were too small to see. I solved both problems by aggregating areas of pixels of size xy to an average value between 0 and 1 and plotting the resulting resized matrix on a scale of red-white-blue. Because the data had already been hierarchically clustered there is not a huge amount of disturbance. This change allows very large datasets to be visualised and that they are still interpretable.