calour
calour copied to clipboard
Robust quantiles
Would be nice if the heatmap colors were scaled to robust quantiles rather than the extreme values
See seaborn robust parameter http://seaborn.pydata.org/generated/seaborn.heatmap.html
second that. looks very useful
Its honestly the only reason why I use seaborn anymore. Is there any interest in adding this into calour?
how much code is it? the main obstacle is dev time
looks like in seaborn what they do is: if vmin is None: vmin = np.percentile(calc_data, 2) if robust else calc_data.min() if vmax is None: vmax = np.percentile(calc_data, 98) if robust else calc_data.max()
so should be a piece of tasty cake :) (main question is how to integrate into the calour heatmap api - what parameter...)
seaborn is just doing robust=True for their parameter I think this should be an easy first pass.
Matplotlib also has vmin/vmax parameters, where you can specify the specific quantiles.
P.S. Is midpoint normalize implemented in calour? https://matplotlib.org/users/colormapnorms.html Seaborn seems to have multiple bugs popping up recently and now it can't perform midpoint normalize anymore -- may issue a bug report in the near future.
On Tue, Jun 25, 2019 at 8:59 AM amnona [email protected] wrote:
looks like in seaborn what they do is: if vmin is None: vmin = np.percentile(calc_data, 2) if robust else calc_data.min() if vmax is None: vmax = np.percentile(calc_data, 98) if robust else calc_data.max()
so should be a piece of tasty cake :) (main question is how to integrate into the calour heatmap api - what parameter...)
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/biocore/calour/issues/38?email_source=notifications&email_token=AA75VXMZ47A2CBKDO7UAHTDP4IJBJA5CNFSM4DSTQXRKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODYQEYOA#issuecomment-505433144, or mute the thread https://github.com/notifications/unsubscribe-auth/AA75VXIPIA7DQEZCIE6ZTS3P4IJBJANCNFSM4DSTQXRA .