Search code examples
rhistogramscatter-plotstan

Is there an R function that can help me modify the size of scatterplot-points in a paris plot?

I used pairs(fit, pars = c("mu", "sigma", "lambda"), include = TRUE) to create a histograms and scatterplots between each pair of variables of my stan model. Right now the points in the scatterplot are very big, like this: enter image description here

I want to make the points smaller and clearer, also if it's possible I would also like to change the color of the plot. Is there an R function that can help me do that? Thank you!

Solution

  • If you just want to plot the points, it can be done by specifying a custom function:

    library(rstan)
example(read_stan_csv)
pairs(fit, pars = c("mu", "sigma", "alpha"), log = TRUE,
panel=function(x,y)points(x,y,col="blue",pch=20,cex=0.6))

    enter image description here

    Under the panel parameter, we specified a point function that has colour "blue", and cex is the parameter to tuning the size of the points.

    The default plot is a smoothScatter plot, instead of individual points, it shows the density of your data points. To change it, you have to call smoothScatter as the panel function, and the colors are provided as a color palette. For example:

    library(RColorBrewer)
PAL = colorRampPalette(c("white",brewer.pal(6,"Greens")))

pairs(fit, pars = c("mu", "sigma", "alpha"), log = TRUE,
panel=function(x,y){smoothScatter(x,y,add=T,colramp = PAL,cex=2)})

    enter image description here

    As mentioned before, it is not a scatterplot so you can see cex has almost no effect. Reading the source code, note there might be some instances where this function above will differ in the number of points plotted.. So you should just specify what you have plotted with the function above and not assume it is equivalent to the pair.stanfit() function.