Why this double x-axis repeats twice?

Question

My problem is related to this other question How to use factorplot to annotate bars with categorical values or to plot 4 variables?

I managed to plot the bars together and put a second axis. But it only works when there are 2 values in Clf column. With 4 values, the x-axis repeats twice. So instead of having 5 groups of 4 bars I get 10 groups of 2 bars? Please check the image to understand the question.

I think the problem is in this lines of code:

pos = np.arange(0,l) % (l//2) + (np.arange(0,l)//(l//2)-1)*0.4
ax.bar(pos, df["Max_Acc"], width=0.4, align="edge", ec="k", color=colors)
handles=[plt.Rectangle((0,0),1,1, color=palette[i], ec="k") for i in range(len(uelec))]

But I really do not know how to fix it.

#CODE FOR THE DATAFRAME
raw_data = {'Max_Acc': [90.71, 87.98, 92.62, 78.93, 73.69,
                       92.62, 94.17, 92.62, 83.81, 79.76,
                       85.36, 89.23, 88.88, 56.23, 89.54,
                       36.25, 14.52, 85.45, 75.45, 45.54],
            'Stage': ['AWA', 'Rem', 'S1', 'S2', 'SWS', 
                      'AWA', 'Rem', 'S1', 'S2', 'SWS',
                      'AWA', 'Rem', 'S1', 'S2', 'SWS', 
                      'AWA', 'Rem', 'S1', 'S2', 'SWS'],
            'Elec': ['Fp1', 'Fp2', 'C4', 'Cz', 'Pz',
                  'C4', 'T3', 'Fp1', 'P4', 'Fp2',
                  'Cz', 'P3', 'P4', 'T4', 'Cp1', 
                  'Cp2', 'T3', 'T4', 'Fp2', 'Fp1'],
            'Clf': ['RF', 'RF', 'RF', 'RF', 'RF',
                    'XG', 'XG', 'XG', 'XG', 'XG', 
                    'SVM','SVM','SVM','SVM','SVM',
                    'KNN','KNN','KNN','KNN','KNN']}

df=pd.DataFrame(raw_data, columns = ['Max_Acc', 'Stage', 'Elec', 'Clf'])
df


#CODE FOR THE PLOT
#To use seaborn palette
palette = sns.color_palette("Set1", 12)
sns.set(style="white")

uelec, uind = np.unique(df["Elec"], return_inverse=1)
cmap = plt.cm.get_cmap("Set1")

colors= [ palette[i] for i in uind]
fig, ax=plt.subplots(figsize=(15, 5)) 
l = len(df)
pos = np.arange(0,l) % (l//2) + (np.arange(0,l)//(l//2)-1)*0.4

ax.bar(pos, df["Max_Acc"], width=0.4, align="edge", ec="k", color=colors)

handles=[plt.Rectangle((0,0),1,1, color=palette[i], ec="k") for i in range(len(uelec))]

legend=ax.legend(bbox_to_anchor=(0., 1.15, 1., .102), handles=handles, labels=list(uelec),
   prop ={'size':10}, loc=9, ncol=8, title=r'Best algorithm using Max_Acc after undersampling' )

legend.get_frame().set_linewidth(0.0) 
plt.setp(legend.get_title(),fontsize='24')

ax.set_xticks(range(l//2))
ax.set_xticklabels(df["Stage"][:l//2])
ax.set_ylim(0, 110)
ax.get_yaxis().set_visible(False)
ax.spines['top'].set_visible(False) 

#Double x-axis
ax.set_xticks(pos+0.2, minor=True)
clf=df['Clf'].tolist()   
ax.set_xticklabels(clf, minor=True)
plt.setp(ax.get_xticklabels(), rotation=0)
ax.tick_params(axis='x', which='major', pad=25, size=0)

ax=ax 
def annotateBars(row, ax=ax): 
    for p in ax.patches:
        ax.annotate("%.2f" % p.get_height(), (p.get_x() + p.get_width() / 2., p.get_height()),
             ha='center', va='center', fontsize=11, color='gray', rotation=90, xytext=(0, 20),
             textcoords='offset points')  

plot = df.apply(annotateBars, ax=ax, axis=1)

IMAGE I GET

IMAGE IT SHOULD BE:

You forgot to tell how the plot is produced from the given dataframe. — ImportanceOfBeingErnest, Oct 19 '17 at 12:49
If you are using the code in your other (linked) question, I think the issue is with your `pos` array. Print that out, and have a think about what it should be... at the moment that is split into two halves, which I guess will lead to the issue you have here — tmdavison, Oct 19 '17 at 13:00

score 1 · Accepted Answer · answered Oct 20 '17 at 11:18

To plot a grouped barplot with multiple labels you may refer to this question. In order to be able to simply plot the dataframe as grouped bar plot, you may pivot your dataframe, once for the values and once for the algorithms. You can then create a unique mapping of algorithm names to numbers which can be applied by a colormap. Best choose a colormap with more than 11 different colors.

import matplotlib.pyplot as plt
import pandas as pd

#CODE FOR THE DATAFRAME
raw_data = {'Max_Acc': [90.71, 87.98, 92.62, 78.93, 73.69,
                       92.62, 94.17, 92.62, 83.81, 79.76,
                       85.36, 89.23, 88.88, 56.23, 89.54,
                       36.25, 14.52, 85.45, 75.45, 45.54],
            'Stage': ['AWA', 'Rem', 'S1', 'S2', 'SWS', 
                      'AWA', 'Rem', 'S1', 'S2', 'SWS',
                      'AWA', 'Rem', 'S1', 'S2', 'SWS', 
                      'AWA', 'Rem', 'S1', 'S2', 'SWS'],
            'Elec': ['Fp1', 'Fp2', 'C4', 'Cz', 'Pz',
                  'C4', 'T3', 'Fp1', 'P4', 'Fp2',
                  'Cz', 'P3', 'P4', 'T4', 'Cp1', 
                  'Cp2', 'T3', 'T4', 'Fp2', 'Fp1'],
            'Clf': ['RF', 'RF', 'RF', 'RF', 'RF',
                    'XG', 'XG', 'XG', 'XG', 'XG', 
                    'SVM','SVM','SVM','SVM','SVM',
                    'KNN','KNN','KNN','KNN','KNN']}

df=pd.DataFrame(raw_data, columns = ['Max_Acc', 'Stage', 'Elec', 'Clf'])

# pivot the dataframe for values and algorithims
df2= pd.pivot(df["Stage"], df['Clf'], df['Max_Acc'])
df3= pd.pivot(df["Stage"], df['Clf'], df['Elec'])

# maping between unique algorithms and some numbers
mapping = dict(zip(df['Elec'].unique(), range(len(df['Elec'].unique()))))
df4 = df3.applymap(lambda x: mapping[x])
palette = plt.cm.tab20


# plot the grouped barplot
ax = df2.plot(kind="bar")

# create positions, colorize bars and annotate bars
pos = []
for (bar, ind) in zip(ax.patches, df4.values.T.flatten()):
    pos.append(bar.get_x()+bar.get_width()/2.)
    bar.set_color(palette(ind))
    ax.annotate("%.2f"%bar.get_height(), (bar.get_x()+bar.get_width()/2., 
                                            bar.get_height()),
             ha='center', va='center', fontsize=10, color='gray', 
             rotation=90, xytext=(0, 20), textcoords='offset points')

# Label minor ticks
ax.set_xticks(pos,minor=True)
lab = []
for i in range(len(pos)):
    l = df2.columns.values[i//len(df2.index.values)]
    lab.append(l)

ax.set_xticklabels(lab,minor=True, rotation=90)
ax.tick_params(axis='x', which='major', pad=30, size=0)
plt.setp(ax.get_xticklabels(), rotation=0)

# legend
handle = lambda ind : plt.Rectangle([0,0],0,0,color=palette(ind))
h, l = zip(*[(handle(ind), la) for (la, ind) in mapping.items() ])

legend=ax.legend(handles=h, labels=l, bbox_to_anchor=(0., 1.15, 1., .102), 
   prop ={'size':10}, loc=9, ncol=6, columnspacing=1,
   title=r'Best algorithm using Max_Acc after undersampling' )

# layout
plt.tight_layout()
plt.subplots_adjust(top=0.8)

ax.set_ylim(0, 120)
ax.get_yaxis().set_visible(False)
ax.spines['top'].set_visible(False) 

plt.show()

Thank you. But do you know how can I make the groups of bars closer? — Aizzaac, Oct 20 '17 at 15:56
`pandas.plot.bar()` has a `width` parameter that you can use. I.e. `width=1` will leave no space between groups. `width=0.5` should be the default. If you choose something in between, spacing is reduced. — ImportanceOfBeingErnest, Oct 20 '17 at 16:06
Ok. I put it in this line: `ax = df2.plot(kind="bar", width=0.8` — Aizzaac, Oct 20 '17 at 16:22

Why this double x-axis repeats twice?

1 Answers1