Hi everyone,
I am trying to plot some MC samples with GetDist and I was wondering if anybody here could kindly suggest how to overcome a binning problem.
My samples are in the form (log10_x, log10_y). I would like to produce plots (with 1 and 2 sigma contours) in terms of (x, y). My strategy was to first transform the samples (log10_x, log10_y) -> (x, y), then feed those to MCSamples() and use log scales for the axis with get_axes. However, this is not enough, as I would also need to set the binning in the plots to, for instance for a 2D plot:
b = [np.logspace(np.log10(xmin),np.log10(xmax), 50), np.logspace(np.log10(ymin),np.log10(ymax), 50)]
In matplotlib one can do: plt.hist2d(x,y, bins=b), but I cannot figure out how to set this in GetDist. Any suggestion?
Thanks!
GetDist: plots with logarithmic binning?
-
- Posts: 2
- Joined: February 09 2022
- Affiliation: CERN
-
- Posts: 1944
- Joined: September 23 2004
- Affiliation: University of Sussex
- Contact:
Re: GetDist: plots with logarithmic binning?
I normally just plot the log (e.g. log A_s in the Planck results).
You could plot as normal for log10_x, log10_y and then manually change the axis and tick labels using standard matplotlib.
But you are welcome to make a pull request supporting log mapping internally. But you can't transform to x, y - then the sampling density would be too variable for the plot to work well because the binning internally is regular linear so it can use FFTs.
You could plot as normal for log10_x, log10_y and then manually change the axis and tick labels using standard matplotlib.
But you are welcome to make a pull request supporting log mapping internally. But you can't transform to x, y - then the sampling density would be too variable for the plot to work well because the binning internally is regular linear so it can use FFTs.
-
- Posts: 2
- Joined: February 09 2022
- Affiliation: CERN
Re: GetDist: plots with logarithmic binning?
Ok, thank you for the reply!
Best
Best