3D plot with categorical axis [Python/Matplotlib]

Question

Trying to plot a 3 dimensional graph with:

x axis - Values (float) y axis - Values (float) z axis - Category (string)

Now I've tried to convert the z vector using pandas.factorize(table.zcolumn)

Output: (array([ 0,  0,  0, ..., -1, -1,  1]), Index([u'London', u'National'], dtype='object'))

So I can plot the numbers no problem.

You'll see there are NaN values which convert to -1, so when I plot the graph there are a bunch of values at -1. The data hold London, National and NaN categories.

How can I label the axes to fit my data? I feel like there should be a simple function to match it.

On the z-axis I need to reassign the ticks -1 to become 'NA', 0 to become 'London' and 1 to become 'National'

I'd also be interested in a way for doing this with large numbers of categories, so code that does not need manually inputting each category string

regions = pandas.factorize(dataTable.Region[id_range])
regions_num = regions[0]

fig = plot.figure()
ax = fig.add_subplot(111,projection='3d')
ax.scatter(y, x, zs=regions_num)

ax.axes.set_zticklabels(["London","National","N/A"])

plot.show()

Answer 1

You just need to set the zticks to the three z-values corresponding to your categories:

ax.axes.set_zticks(regions_num)

Having said that, I don't this is actually a very good way of plotting your data. 3D plots are most useful when your X, Y and Z values are all continuous variables. Representing regions as different z-levels might make a bit more sense if 'region' was an ordinal variable, but is there any reason why 'N/A' should be 'higher' than 'National'? 3D plots are also generally harder to read than 2D plots - for example, because of the perspective projection a point that's nearby in the 'National' category might look a lot like a point that's further away but in the 'N/A' category.

A more appropriate choice might be to represent these data as a scatter plot on 2D axes, with different colours corresponding to the different categories.