I am not aware of a free plotting package (e.g., gnuplot)
in the Linux distribution that plots y(x) with an *individually
adjustable size for the data points*.
One of the reasons I invented `vplot` was so that
I could make figures like the following:

An attributes diagram showing performance of forecasted probability of clear skies at 10 AM PST, with predictors based on observations at 4 AM PST. The actual observed frequency of occurrence is then plotted against the forecasted probability. The area of the dot is made proportional to number of forecasts at that probability. The frequency of occurrence for the entire dataset is about 0.68 for these events. A horizontal "no resolution" line is drawn at the climatological frequency. A vertical line is also drawn for a forecasted probability at the observed frequency. A "perfect reliability line" is drawn with a slope of 1, and a "no skill" line with half that slope. |

Here is the script to make the attributes diagram.