Depends: data.table and Rmpfr.
Features:
- Over 10x faster that plotting all the points.
- Support p-values less than default floating point limit of ~1e-323.
IMPORTANT: Used properly, this code will result in a QQ plot that is identical to one using all the data points. However, choosing too small a subsample will produce a QQ plot that is an approximation to that of using all data points, although likely a very close approximation.
Usage:
plot_fqq.R <txt> <p.m> <p.e>
Where arguments are:
<txt>: Text file containing a column labeled P with p-values. For fast loading have the file contain only this column.<p.m>: The proportion of points to subsample from the QQ plot. Default to sub-sample 0.1% of p-value distribution, excluding the tails (see next point).<p.e>: Cutoff to select points as being in the tails to plot. Default to include the top and bottom 1% of the p-value distribution into the plot.
Example command:
plot_fqq.R p.txt 0.001 0.01
Example data:
> write.table(data.frame(P=runif(2000000,0,1)), file="p.txt", row.names=FALSE, quote=FALSE)