Currently, scatterplots just project the min/max values to the edges of the available canvas. In some cases that you get plots that can be interpreted:
(mental note: implement labeling individual columns when plotting categorical attributes)
But it looks awkward if the ratio of the attributes doesn't match the ratio of the canvas:
In order to fix this, I propose we add a "lock x/y ratio" option. What this would to is set the ratio of the plot to the min/max delta of each attribute.
So let's say tSNE1 has values between -2 and +3, and tSNE2 has values between -0.5 and +1, then the ratio is 10:3, and when selecting lock ratio the plot is scaled to make it fit that ratio, while maximised for available visible canvas space.
Especially for default clustering attributes¹ this makes sense, since they are supposed to match each other in magnitude.
¹ That is to say: X and _Y, _tSNE1 and _tSNE2, _PC1 and _PC2, _LogMean and _LogCV*.
Currently, scatterplots just project the min/max values to the edges of the available canvas. In some cases that you get plots that can be interpreted:
(mental note: implement labeling individual columns when plotting categorical attributes)
But it looks awkward if the ratio of the attributes doesn't match the ratio of the canvas:
In order to fix this, I propose we add a "lock x/y ratio" option. What this would to is set the ratio of the plot to the min/max delta of each attribute.
So let's say
tSNE1has values between -2 and +3, andtSNE2has values between -0.5 and +1, then the ratio is 10:3, and when selectinglock ratiothe plot is scaled to make it fit that ratio, while maximised for available visible canvas space.Especially for default clustering attributes¹ this makes sense, since they are supposed to match each other in magnitude.
¹ That is to say: X and _Y, _tSNE1 and _tSNE2, _PC1 and _PC2, _LogMean and _LogCV*.