Develop Reference

Matplotlib - duplicated axes - prevent gridlines from covering data - set right ylabel

I want to duplicate axes so that I can express an exponent in terms of its doubling time.
I think I am doing things right, but I have two problems
no label on the right hand side of the chart and
y-axis gridlines that are plotted above the data that I cannot shift to the bottom, nor remove.
Example code follows:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
MARGINS = 0.02
data = pd.Series(np.arange(0.05, 1.0, 0.1))
# preliminaries
plt.style.use('ggplot')
fig, ax = plt.subplots()
ax.figure.set_size_inches((8, 4))
ax.set_ylabel('$k$') # This works
ax.margins(MARGINS)
ax.set_axisbelow(True)
# duplicate the axes
axr = ax.twinx().twiny()
axr.margins(MARGINS)
axr.set_ylabel('Doubling time') # This does not work
# No x-ticks at the top
axr.xaxis.set_ticks([])
axr.xaxis.set_ticklabels([])
# plot the data
ax.plot(data.index, data)
# label right-hand y-axis
locations = ax.get_yticks()
new_labels = [f'{np.log(2)/x:,.2f}' if x != 0 else '∞' for x in locations ]
axr.yaxis.set_ticks(locations)
axr.yaxis.set_ticklabels(new_labels)
axr.set_axisbelow(True) # this does not work
# match the left and right ylim settings
axr.set_ylim(ax.get_ylim())
axr.set_xlim(ax.get_xlim())
# remove the grid
axr.grid(False, which='both')
axr.yaxis.grid(False, which='both') # this does not work
# finish-up
ax.set_title('Chart')
fig.tight_layout(pad=1.1)
plt.show()
plt.close('all')
Desired output:
Similar chart to above but with:
a right hand side y-axis label
no y-axis gridlines over the data line (but keep the horizontal gridlines under the dataline)

Change the order of twiny and twinx:
axr = ax.twiny().twinx()

Categorical bubble plot in Python

I have a dataset with a lot of categorical variables and a binary target variable. What package is available in Python or other opensource GUI-based software where I can scatterplot two categorical variables on the X and Y axis and use the target variable as hue?
I have looked at Seaborn's catplot, but for that, one axis has to be numerical while the other categorical. So it doesn't serve this case.
For example, you can use the following:
import seaborn as sns
data = sns.load_dataset('titanic')
Here are the plot features I want
X-axis - 'embark_town'
Y-axis - 'class'
hue - 'alive'

I am of the opinion that if you have to rearrange a seaborn graph substantially, you can also create this graph from scratch with matplotlib. This gives us the opportunity to have a different approach to display this categorical vs categorical plot:
import matplotlib.pyplot as plt
from matplotlib.markers import MarkerStyle
import numpy as np
#dataframe and categories
import seaborn as sns
df = sns.load_dataset('titanic')
X = "embark_town"
Y = "class"
H = "alive"
bin_dic = {0: "yes", 1: "no"}
#counting the X-Y-H category entries
plt_df = df.groupby([X, Y, H]).size().to_frame(name="vals").reset_index()
#figure preparation with grid and scaling
fig, ax = plt.subplots(figsize=(9, 6))
ax.set_ylim(plt_df[Y].unique().size-0.5, -0.5)
ax.set_xlim(-0.5, plt_df[X].unique().size+1.0)
ax.grid(ls="--")
#upscale factor for scatter marker size
scale=10000/plt_df.vals.max()
#left marker for category 0
ax.scatter(plt_df[plt_df[H]==bin_dic[0]][X],
plt_df[plt_df[H]==bin_dic[0]][Y],
s=plt_df[plt_df[H]==bin_dic[0]].vals*scale,
c=[(0, 0, 1, 0.5)], edgecolor="black", marker=MarkerStyle("o", fillstyle="left"),
label=bin_dic[0])
#right marker for category 1
ax.scatter(plt_df[plt_df[H]==bin_dic[1]][X],
plt_df[plt_df[H]==bin_dic[1]][Y],
s=plt_df[plt_df[H]==bin_dic[1]].vals*scale,
c=[(1, 0, 0, 0.5)], edgecolor="black", marker=MarkerStyle("o", fillstyle="right"),
label=bin_dic[1])
#legend entries for the two categories
l = ax.legend(title="Survived the catastrophe", ncol=2, framealpha=0, loc="upper right", columnspacing=0.1,labelspacing=1.5)
l.legendHandles[0]._sizes = l.legendHandles[1]._sizes = [800]
#legend entries representing sizes
bubbles_n=5
bubbles_min = 50*(1+plt_df.vals.min()//50)
bubbles_step = 10*((plt_df.vals.max()-bubbles_min)//(10*(bubbles_n-1)))
bubbles_x = plt_df[X].unique().size+0.5
for i, bubbles_y in enumerate(np.linspace(0.5, plt_df[Y].unique().size-1, bubbles_n)):
#plot each legend bubble to indicate different marker sizes
ax.scatter(bubbles_x,
bubbles_y,
s=(bubbles_min + i*bubbles_step) * scale,
c=[(1, 0, 1, 0.6)], edgecolor="black")
#and label it with a value
ax.annotate(bubbles_min+i*bubbles_step, xy=(bubbles_x, bubbles_y),
ha="center", va="center",
fontsize="large", fontweight="bold", color="white")
plt.show()

Seaborn supports, just like matplotlib, the plotting of categorical vs categorical variables. One can create semitransparent markers that allow to see both categories, although this might be difficult to distinguish from one marker if both are of similar size. The essential plot is rather easy - we transform the dataframe with groupby and size to count the entries per triplet embarking town - class - alive category, then create a scatterplot with count value as markersize. However, the legend entry is the complicated part here. Either the markersize is tiny in the plot or massive in the legend. I tried to balance this but I am not happy with the result. A lot of manual adjusting necessary here, so seaborn is no real advantage here. Any suggestions on how to simplify this within seaborn are welcome.
import seaborn as sns
import matplotlib.pyplot as plt
#dataframe and categories
df = sns.load_dataset('titanic')
X = "embark_town"
Y = "class"
H = "alive"
#counting the X-Y-H category entries
plt_df = df.groupby([X, Y, H]).size().to_frame(name="people").reset_index()
#figure preparation with grid and scaling
fig, ax = plt.subplots(figsize=(6,4))
ax.set_ylim(plt_df[Y].unique().size-0.5, -0.5)
ax.set_xlim(-0.5, plt_df[X].unique().size+1.0)
ax.grid(ls="--")
#the actual scatterplot with markersize representing the counted values
sns.scatterplot(x=X,
y=Y,
size="people",
sizes=(100, 10000),
alpha=0.5,
edgecolor="black",
hue=H,
data=plt_df,
ax=ax)
#creating two legends because the hue markers differ in size from the others
handles, labels = ax.get_legend_handles_labels()
l = ax.legend(handles[:3], labels[:3], title="The poor die first", markerscale=2, loc="upper right")
ax.add_artist(l)
#and seaborn plots the size markers in black, so you would get massive black blobs in the legend
#we change the color and make them transparent
for handle in handles:
handle.set_facecolors((0, 1, 1, 0.5))
ax.legend(handles[4::2], labels[4::2], title="N° of people", loc="lower right", handletextpad=4, labelspacing=3, markerfirst=False)
plt.tight_layout()
plt.show()
Sample output:

How to combine 2 dataframe histograms in 1 plot?

I would like to use a code that shows all histograms in a dataframe. That will be df.hist(bins=10). However, I would like to add another histograms which shows CDF df_hist=df.hist(cumulative=True,bins=100,density=1,histtype="step")
I tried separating their matplotlib axes by using fig=plt.figure() and
plt.subplot(211). But this df.hist is actually part of pandas function, not matplotlib function. I also tried setting axes and adding ax=ax1 and ax2 options to each histogram but it didn't work.
How can I combine these histograms together?
Any help?
Histograms that I want to combine are like these. I want to show them side by side or put the second one on tip of the first one.
Sorry that I didn't care to make them look good.

It is possible to draw them together:
# toy data frame
df = pd.DataFrame(np.random.normal(0,1,(100,20)))
# draw hist
fig, axes = plt.subplots(5,4, figsize=(16,10))
df.plot(kind='hist', subplots=True, ax=axes, alpha=0.5)
# clone axes so they have different scales
ax_new = [ax.twinx() for ax in axes.flatten()]
df.plot(kind='kde', ax=ax_new, subplots=True)
plt.show()
Output:
It's also possible to draw them side-by-side. For example
fig, axes = plt.subplots(10,4, figsize=(16,10))
hist_axes = axes.flatten()[:20]
df.plot(kind='hist', subplots=True, ax=hist_axes, alpha=0.5)
kde_axes = axes.flatten()[20:]
df.plot(kind='kde', subplots=True, ax=kde_axes, alpha=0.5)
will plot hist on top of kde.

You can find more info here: Multiple histograms in Pandas (possible duplicate btw) but apparently Pandas cannot handle multiple histogram on same graphs.
It's ok because np.histogram and matplotlib.pyplot can, check the above link for a more complete answer.

Solution for overlapping histograms with df.hist with any number of subplots
You can combine two dataframe histogram figures by creating twin axes using the grid of axes returned by df.hist. Here is an example of normal histograms combined with cumulative step histograms where the size of the figure and the layout of the grid of subplots are taken care of automatically:
import numpy as np # v 1.19.2
import pandas as pd # v 1.1.3
import matplotlib.pyplot as plt # v 3.3.2
# Create sample dataset stored in a pandas dataframe
rng = np.random.default_rng(seed=1) # random number generator
letters = [chr(i) for i in range(ord('A'), ord('G')+1)]
df = pd.DataFrame(rng.exponential(1, size=(100, len(letters))), columns=letters)
# Set parameters for figure dimensions and grid layout
nplots = df.columns.size
ncols = 3
nrows = int(np.ceil(nplots/ncols))
subp_w = 10/ncols # 10 is the total figure width in inches
subp_h = 0.75*subp_w
bins = 10
# Plot grid of histograms with pandas function (with a shared y-axis)
grid = df.hist(grid=False, sharey=True, figsize=(ncols*subp_w, nrows*subp_h),
layout=(nrows, ncols), bins=bins, edgecolor='white', linewidth=0.5)
# Create list of twin axes containing second y-axis: note that due to the
# layout, the grid object may contain extra unused axes that are not shown
# (here in the H and I positions). The ax parameter of df.hist only accepts
# a number of axes that corresponds to the number of numerical variables
# in df, which is why the flattened array of grid axes is sliced here.
grid_twinx = [ax.twinx() for ax in grid.flat[:nplots]]
# Plot cumulative step histograms over normal histograms: note that the grid layout is
# preserved in grid_twinx so no need to set the layout parameter a second time here.
df.hist(ax=grid_twinx, histtype='step', bins=bins, cumulative=True, density=True,
color='tab:orange', linewidth=2, grid=False)
# Adjust space between subplots after generating twin axes
plt.gcf().subplots_adjust(wspace=0.4, hspace=0.4)
plt.show()
Solution for displaying histograms of different types side-by-side with matplotlib
To my knowledge, it is not possible to show the different types of plots side-by-side with df.hist. You need to create the figure from scratch, like in this example using the same dataset as before:
# Set parameters for figure dimensions and grid layout
nvars = df.columns.size
plot_types = 2 # normal histogram and cumulative step histogram
ncols_vars = 2
nrows = int(np.ceil(nvars/ncols_vars))
subp_w = 10/(plot_types*ncols_vars) # 10 is the total figure width in inches
subp_h = 0.75*subp_w
bins = 10
# Create figure with appropriate size
fig = plt.figure(figsize=(plot_types*ncols_vars*subp_w, nrows*subp_h))
fig.subplots_adjust(wspace=0.4, hspace=0.7)
# Create subplots by adding a new axes per type of plot for each variable
# and create lists of axes of normal histograms and their y-axis limits
axs_hist = []
axs_hist_ylims = []
for idx, var in enumerate(df.columns):
axh = fig.add_subplot(nrows, plot_types*ncols_vars, idx*plot_types+1)
axh.hist(df[var], bins=bins, edgecolor='white', linewidth=0.5)
axh.set_title(f'{var} - Histogram', size=11)
axs_hist.append(axh)
axs_hist_ylims.append(axh.get_ylim())
axc = fig.add_subplot(nrows, plot_types*ncols_vars, idx*plot_types+2)
axc.hist(df[var], bins=bins, density=True, cumulative=True,
histtype='step', color='tab:orange', linewidth=2)
axc.set_title(f'{var} - Cumulative step hist.', size=11)
# Set shared y-axis for histograms
for ax in axs_hist:
ax.set_ylim(max(axs_hist_ylims))
plt.show()

Creating a percentage countplot in python with pandas [duplicate]

I have a Pandas DataFrame with a column called "AXLES", which can take an integer value between 3-12. I am trying to use Seaborn's countplot() option to achieve the following plot:
left y axis shows the frequencies of these values occurring in the data. The axis extends are [0%-100%], tick marks at every 10%.
right y axis shows the actual counts, values correspond to tick marks determined by the left y axis (marked at every 10%.)
x axis shows the categories for the bar plots [3, 4, 5, 6, 7, 8, 9, 10, 11, 12].
Annotation on top of the bars show the actual percentage of that category.
The following code gives me the plot below, with actual counts, but I could not find a way to convert them into frequencies. I can get the frequencies using df.AXLES.value_counts()/len(df.index) but I am not sure about how to plug this information into Seaborn's countplot().
I also found a workaround for the annotations, but I am not sure if that is the best implementation.
Any help would be appreciated!
Thanks
plt.figure(figsize=(12,8))
ax = sns.countplot(x="AXLES", data=dfWIM, order=[3,4,5,6,7,8,9,10,11,12])
plt.title('Distribution of Truck Configurations')
plt.xlabel('Number of Axles')
plt.ylabel('Frequency [%]')
for p in ax.patches:
ax.annotate('%{:.1f}'.format(p.get_height()), (p.get_x()+0.1, p.get_height()+50))
EDIT:
I got closer to what I need with the following code, using Pandas' bar plot, ditching Seaborn. Feels like I'm using so many workarounds, and there has to be an easier way to do it. The issues with this approach:
There is no order keyword in Pandas' bar plot function as Seaborn's countplot() has, so I cannot plot all categories from 3-12 as I did in the countplot(). I need to have them shown even if there is no data in that category.
The secondary y-axis messes up the bars and the annotation for some reason (see the white gridlines drawn over the text and bars).
plt.figure(figsize=(12,8))
plt.title('Distribution of Truck Configurations')
plt.xlabel('Number of Axles')
plt.ylabel('Frequency [%]')
ax = (dfWIM.AXLES.value_counts()/len(df)*100).sort_index().plot(kind="bar", rot=0)
ax.set_yticks(np.arange(0, 110, 10))
ax2 = ax.twinx()
ax2.set_yticks(np.arange(0, 110, 10)*len(df)/100)
for p in ax.patches:
ax.annotate('{:.2f}%'.format(p.get_height()), (p.get_x()+0.15, p.get_height()+1))

You can do this by making a twinx axes for the frequencies. You can switch the two y axes around so the frequencies stay on the left and the counts on the right, but without having to recalculate the counts axis (here we use tick_left() and tick_right() to move the ticks and set_label_position to move the axis labels
You can then set the ticks using the matplotlib.ticker module, specifically ticker.MultipleLocator and ticker.LinearLocator.
As for your annotations, you can get the x and y locations for all 4 corners of the bar with patch.get_bbox().get_points(). This, along with setting the horizontal and vertical alignment correctly, means you don't need to add any arbitrary offsets to the annotation location.
Finally, you need to turn the grid off for the twinned axis, to prevent grid lines showing up on top of the bars (ax2.grid(None))
Here is a working script:
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
import matplotlib.ticker as ticker
# Some random data
dfWIM = pd.DataFrame({'AXLES': np.random.normal(8, 2, 5000).astype(int)})
ncount = len(dfWIM)
plt.figure(figsize=(12,8))
ax = sns.countplot(x="AXLES", data=dfWIM, order=[3,4,5,6,7,8,9,10,11,12])
plt.title('Distribution of Truck Configurations')
plt.xlabel('Number of Axles')
# Make twin axis
ax2=ax.twinx()
# Switch so count axis is on right, frequency on left
ax2.yaxis.tick_left()
ax.yaxis.tick_right()
# Also switch the labels over
ax.yaxis.set_label_position('right')
ax2.yaxis.set_label_position('left')
ax2.set_ylabel('Frequency [%]')
for p in ax.patches:
x=p.get_bbox().get_points()[:,0]
y=p.get_bbox().get_points()[1,1]
ax.annotate('{:.1f}%'.format(100.*y/ncount), (x.mean(), y),
ha='center', va='bottom') # set the alignment of the text
# Use a LinearLocator to ensure the correct number of ticks
ax.yaxis.set_major_locator(ticker.LinearLocator(11))
# Fix the frequency range to 0-100
ax2.set_ylim(0,100)
ax.set_ylim(0,ncount)
# And use a MultipleLocator to ensure a tick spacing of 10
ax2.yaxis.set_major_locator(ticker.MultipleLocator(10))
# Need to turn the grid on ax2 off, otherwise the gridlines end up on top of the bars
ax2.grid(None)
plt.savefig('snscounter.pdf')

I got it to work using core matplotlib's bar plot. I didn't have your data obviously, but adapting it to yours should be straight forward.
Approach
I used matplotlib's twin axis and plotted the data as bars on the second Axes object. The rest ist just some fiddeling around to get the ticks right and make annotations.
Hope this helps.
Code
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib
from mpl_toolkits.mplot3d import Axes3D
import seaborn as sns
tot = np.random.rand( 1 ) * 100
data = np.random.rand( 1, 12 )
data = data / sum(data,1) * tot
df = pd.DataFrame( data )
palette = sns.husl_palette(9, s=0.7 )
### Left Axis
# Plot nothing here, autmatically scales to second axis.
fig, ax1 = plt.subplots()
ax1.set_ylim( [0,100] )
# Remove grid lines.
ax1.grid( False )
# Set ticks and add percentage sign.
ax1.yaxis.set_ticks( np.arange(0,101,10) )
fmt = '%.0f%%'
yticks = matplotlib.ticker.FormatStrFormatter( fmt )
ax1.yaxis.set_major_formatter( yticks )
### Right Axis
# Plot data as bars.
x = np.arange(0,9,1)
ax2 = ax1.twinx()
rects = ax2.bar( x-0.4, np.asarray(df.loc[0,3:]), width=0.8 )
# Set ticks on x-axis and remove grid lines.
ax2.set_xlim( [-0.5,8.5] )
ax2.xaxis.set_ticks( x )
ax2.xaxis.grid( False )
# Set ticks on y-axis in 10% steps.
ax2.set_ylim( [0,tot] )
ax2.yaxis.set_ticks( np.linspace( 0, tot, 11 ) )
# Add labels and change colors.
for i,r in enumerate(rects):
h = r.get_height()
r.set_color( palette[ i % len(palette) ] )
ax2.text( r.get_x() + r.get_width()/2.0, \
h + 0.01*tot, \
r'%d%%'%int(100*h/tot), ha = 'center' )

I think you can first set the y major ticks manually and then modify each label
dfWIM = pd.DataFrame({'AXLES': np.random.randint(3, 10, 1000)})
total = len(dfWIM)*1.
plt.figure(figsize=(12,8))
ax = sns.countplot(x="AXLES", data=dfWIM, order=[3,4,5,6,7,8,9,10,11,12])
plt.title('Distribution of Truck Configurations')
plt.xlabel('Number of Axles')
plt.ylabel('Frequency [%]')
for p in ax.patches:
ax.annotate('{:.1f}%'.format(100*p.get_height()/total), (p.get_x()+0.1, p.get_height()+5))
#put 11 ticks (therefore 10 steps), from 0 to the total number of rows in the dataframe
ax.yaxis.set_ticks(np.linspace(0, total, 11))
#adjust the ticklabel to the desired format, without changing the position of the ticks.
_ = ax.set_yticklabels(map('{:.1f}%'.format, 100*ax.yaxis.get_majorticklocs()/total))

Can I mimic a log scale of an axis in matplotlib without transforming the associated data?

I am trying to display a Zipf plot, which is typically displayed on a log-log scale.
I'm using a library which gives rank in linear scale and frequencies in log scale. I have the following code which plots my data fairly correctly:
ranks = [3541, 60219, 172644, 108926, 733215, 1297533, 1297534, 1297535]
# These frequencies are already log-scale
freqs = [-10.932271003723145, -15.213129043579102, -17.091760635375977, -16.27560806274414,
-19.482173919677734, -19.502029418945312, -19.502029418945312, -19.502029418945312]
data = {
'ranks': ranks,
'freqs': freqs,
}
df = pd.DataFrame(data=data)
_, ax = plt.subplots(figsize=(7, 7))
ax.set(xscale="log", yscale="linear")
ax.set_title("Zipf plot")
sns.regplot("ranks", "freqs", data=df, ax=ax, fit_reg=False)
ax.set_xlabel("Frequency rank of token")
ax.set_ylabel("Absolute frequency of token")
ax.grid(True, which="both")
plt.show()
The resulting plot is:
The plot looks good, but the y-label is weird. I'd like it to be displayed in log-increments as well. My current workaround is to raise 10 to the power of each element in the freqs list; i.e.,
freqs = [10**freq for freq in freqs]
# ...
and change the yscale in ax.set to log; i.e.,
_, ax = plt.subplots(figsize=(7, 7))
ax.set(xscale="log", yscale="log")
ax.set_title("Zipf plot")
# ...
This gives me the expected plot (below), but it requires a transform of the data which is a) relatively expensive, b) redundant, c) lossy.
Is there a way to mimic the log scale of the axes in a matplotlib plot without transforming the data?

First a comment: Personally i would prefer the method of rescaling the data, since it makes everything much easier at the expense of some more memory/cpu time and accurary should not matter
Now to the question, which is acutally how to mimic a log scale on a linear axis
Solution 1: mimic the log scale
This is not easy. Setting the axes to log scale changes a lot in the background and one needs to mimic all of that.
The easy part is to set the major tickmark frequency to 1 by using matplotlib.ticker.MultipleLocator()
Creating the minor tickmarks at positions which look logarithmic is harder. The best solution I could come up with is to set them manually using the matplotlib.ticker.FixedLocator()
Last we need to change the tickmarks to represent the actual numbers, meaning that they should look like 10^(-x) instead of -x. I am aware of two options here:
Using a FuncFormatter that sets the values 10**x in scientific format.
Using a FuncFormatter that sets the values 10^x in Latex format. This looks much nicer but contrasts to the rest of the plot.
I do not know any better solution for that last point, but maybe someone else does.
Here is the code and how it looks.
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import numpy as np
from matplotlib.ticker import MultipleLocator, FixedLocator, FuncFormatter
###### Locators for Y-axis
# set tickmarks at multiples of 1.
majorLocator = MultipleLocator(1.)
# create custom minor ticklabels at logarithmic positions
ra = np.array([ [n+(1.-np.log10(i))] for n in xrange(10,20) for i in [2,3,4,5,6,7,8,9][::-1]]).flatten()*-1.
minorLocator = FixedLocator(ra)
###### Formatter for Y-axis (chose any of the following two)
# show labels as powers of 10 (looks ugly)
majorFormatter= FuncFormatter(lambda x,p: "{:.1e}".format(10**x) )
# or using MathText (looks nice, but not conform to the rest of the layout)
majorFormatter= FuncFormatter(lambda x,p: r"$10^{"+"{x:d}".format(x=int(x))+r"}$" )
ranks = [3541, 60219, 172644, 108926, 733215, 1297533, 1297534, 1297535]
# These frequencies are already log-scale
freqs = [-10.932271003723145, -15.213129043579102, -17.091760635375977, -16.27560806274414,
-19.482173919677734, -19.502029418945312, -19.502029418945312, -19.502029418945312]
data = {
'ranks': ranks,
'freqs': freqs,
}
df = pd.DataFrame(data=data)
_, ax = plt.subplots(figsize=(6, 6))
ax.set(xscale="log", yscale="linear")
ax.set_title("Zipf plot")
sns.regplot("ranks", "freqs", data=df, ax=ax, fit_reg=False)
# Set the locators
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
# Set formatter if you like to have the ticklabels consistently in power notation
ax.yaxis.set_major_formatter(majorFormatter)
ax.set_xlabel("Frequency rank of token")
ax.set_ylabel("Absolute frequency of token")
ax.grid(True, which="both")
plt.show()
Solution 2: Use a different axes
A different solution, of which I did not think in the first place, would be to use two different axes, one with a loglog scale which looks nice and produces the correct labels and ticks and anotherone to plot the data to.
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import numpy as np
ranks = [3541, 60219, 172644, 108926, 733215, 1297533, 1297534, 1297535]
# These frequencies are already log-scale
freqs = [-10.932271003723145, -15.213129043579102, -17.091760635375977, -16.27560806274414,
-19.482173919677734, -19.502029418945312, -19.502029418945312, -19.502029418945312]
data = {
'ranks': ranks,
'freqs': freqs,
}
df = pd.DataFrame(data=data)
fig, ax = plt.subplots(figsize=(6, 6))
# use 2 axes
# ax is the log, log scale which produces nice labels and ticks
ax.set(xscale="log", yscale="log")
ax.set_title("Zipf plot")
# ax2 is the axes where the values are plottet to
ax2 = ax.twinx()
#plot values to ax2
sns.regplot("ranks", "freqs", data=df, ax=ax2, fit_reg=False)
# set the limits of the log log axis to 10 to the power of the label of ax2
ax.set_ylim(10**np.array(ax2.get_ylim()) )
ax.set_xlabel("Frequency rank of token")
ax.set_ylabel("Absolute frequency of token")
# remove ticklabels and axislabel from ax2
ax2.set_yticklabels([])
ax2.set_ylabel("")
ax.grid(True, which="both")
plt.show()