A variety of links that we’ve run across; hopefully these will get incorporated in the primary course material eventually, but in the meantime enjoy browsing!
rdeck package: https://twitter.com/MilesMcBain/status/1448076055651053569ggdist, see, etc. packages: https://twitter.com/mjskay/status/1419792270258225153ggpattern packagehues package (R implementation of IWantHue) now exists on CRANhere are my thoughts about plotting veg composition by period: plotting compositional data is fundamentally difficult. There is a tension between
- making it easy to read off quantitative information (i.e. using Cleveland hierarchy, which says it is easiest to judge magnitudes when we are comparing the position of points along a linear scale, e.g. along the x- or y-axis of a plot)
- expressing the ‘compositionality’ of the data set, i.e. that the sum of the types is always 1 (which would suggest something like stacked bar charts or treemaps - although treemaps are arguably even worse because they represent magnitudes by area rather than by (linear) areas). (Ternary diagrams are a particularly nice approach but they only work when there are exactly three categories.)
- This gets even harder when there are a large number of categories (microbiome, allele frequencies, etc.) - luckily this is not at play here.
My solution here does some more processing to get confidence intervals for the proportions of each type (which I always like to include if possible), based on binomial/multinomial sampling.