Horizontal barplot with offset in seaborn
Question:
My dataset is like this, where the data points in each row or column are pandas objects.
Here is the dataset:
https://github.com/aebk2015/multipleboxplot.git
I want to have bar plots for each of the columns "Location" (P1 -P14) for each categories (92A11, 92B11, 82B11); something like this:
I have tried something like this and i can have a bar plots for each individual Pi (i=1…14) but not only is it a laborious, it does not look what I want:
fig, ax = plt.subplots(2, 3, figsize=(8,2))
sns.stripplot(data=df.loc[7]['92A11'].split(','), dodge=True, linewidth=1, ax=ax[0,0], color='black', jitter=False, orient='h')
sns.violinplot(data=df.loc[7]['92A11'].split(','), ax=ax[0,0], color='orange', orient='h')
sns.stripplot(data=df.loc[7]['92B11'].split(','), dodge=True, linewidth=1, ax=ax[0,1], color='black', jitter=False, orient='h')
sns.violinplot(data=df.loc[7]['92B11'].split(','), ax=ax[0,1], color='orange', orient='h')
sns.stripplot(data=df.loc[7]['82B11'].split(','), dodge=True, linewidth=1, ax=ax[0,2], color='black', jitter=False, orient='h')
sns.violinplot(data=df.loc[7]['82B11'].split(','), ax=ax[0,2], color='orange', orient='h')
sns.stripplot(data=df.loc[6]['92A11'].split(','), dodge=True, linewidth=1, ax=ax[1,0], color='black', jitter=False, orient='h')
sns.violinplot(data=df.loc[6]['92A11'].split(','), ax=ax[1,0], color='orange', orient='h')
sns.stripplot(data=df.loc[6]['92B11'].split(','), dodge=True, linewidth=1, ax=ax[1,1], color='black', jitter=False, orient='h')
sns.violinplot(data=df.loc[6]['92B11'].split(','), ax=ax[1,1], color='orange', orient='h')
sns.stripplot(data=df.loc[6]['82B11'].split(','), dodge=True, linewidth=1, ax=ax[1,2], color='black', jitter=False, orient='h')
sns.violinplot(data=df.loc[6]['82B11'].split(','), ax=ax[1,2], color='orange', orient='h')
ax[0,0].set_xlim(0,200)
ax[0,1].set_xlim(0,200)
ax[0,2].set_xlim(0,200)
ax[1,0].set_xlim(0,200)
ax[1,1].set_xlim(0,200)
ax[1,2].set_xlim(0,200)
ax[1,0].set_xlabel('92A11')
ax[1,1].set_xlabel('92A11')
ax[1,2].set_xlabel('92A11')
ax[0,0].set_ylabel('P8')
ax[1,0].set_ylabel('P7')
fig.tight_layout()
Answers:
Seaborn works easiest with its data in long form. pd.melt can convert the 3 individual columns into a variable-name and value column. pd.explode can convert lists in cells into multiple rows. As the current entries are strings, they can be converted to lists of numbers before calling .explode().
The following code supposes you want box plots, not bars. It also supposes that an entry with only '0' is supposed to stay empty.
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
df = pd.read_csv('raw_data.csv')
df = df.rename(columns={df.columns[0]: 'Location'}) # suppose the first column is named 'Location'
df_melted = df.melt(id_vars='Location', var_name='Dataset', value_name='Datapoint')
df_melted['Datapoint'] = df_melted['Datapoint'].map(
lambda s: [np.nan] if s == '0' else [float(v) for v in s.split(',')])
df_exploded = df_melted.explode(column='Datapoint', ignore_index=True)
df_exploded['Datapoint'] = pd.to_numeric(df_exploded['Datapoint']) # make sure the column is numeric
g = sns.catplot(data=df_exploded, col='Dataset', kind='box', x='Datapoint', y='Location')
g.set(xlabel='', ylabel='') # remove x and y label
PS: df_merged looks like:
Location Dataset Datapoint
0 P1 92A11 [nan]
1 P2 92A11 [10.0, 15.0]
2 P3 92A11 [100.0, 40.0, 60.0, 50.0, 70.0, 50.0, 30.0, 40...
3 P4 92A11 [55.0, 40.0, 20.0, 15.0, 15.0, 15.0, 15.0, 10....
4 P5 92A11 [nan]
...
df_exploded has each value inside the list into a separate row:
Location Dataset Datapoint
0 P1 92A11 NaN
1 P2 92A11 10.0
1 P2 92A11 15.0
2 P3 92A11 100.0
2 P3 92A11 40.0
...
My dataset is like this, where the data points in each row or column are pandas objects.
Here is the dataset:
https://github.com/aebk2015/multipleboxplot.git
I want to have bar plots for each of the columns "Location" (P1 -P14) for each categories (92A11, 92B11, 82B11); something like this:
I have tried something like this and i can have a bar plots for each individual Pi (i=1…14) but not only is it a laborious, it does not look what I want:
fig, ax = plt.subplots(2, 3, figsize=(8,2))
sns.stripplot(data=df.loc[7]['92A11'].split(','), dodge=True, linewidth=1, ax=ax[0,0], color='black', jitter=False, orient='h')
sns.violinplot(data=df.loc[7]['92A11'].split(','), ax=ax[0,0], color='orange', orient='h')
sns.stripplot(data=df.loc[7]['92B11'].split(','), dodge=True, linewidth=1, ax=ax[0,1], color='black', jitter=False, orient='h')
sns.violinplot(data=df.loc[7]['92B11'].split(','), ax=ax[0,1], color='orange', orient='h')
sns.stripplot(data=df.loc[7]['82B11'].split(','), dodge=True, linewidth=1, ax=ax[0,2], color='black', jitter=False, orient='h')
sns.violinplot(data=df.loc[7]['82B11'].split(','), ax=ax[0,2], color='orange', orient='h')
sns.stripplot(data=df.loc[6]['92A11'].split(','), dodge=True, linewidth=1, ax=ax[1,0], color='black', jitter=False, orient='h')
sns.violinplot(data=df.loc[6]['92A11'].split(','), ax=ax[1,0], color='orange', orient='h')
sns.stripplot(data=df.loc[6]['92B11'].split(','), dodge=True, linewidth=1, ax=ax[1,1], color='black', jitter=False, orient='h')
sns.violinplot(data=df.loc[6]['92B11'].split(','), ax=ax[1,1], color='orange', orient='h')
sns.stripplot(data=df.loc[6]['82B11'].split(','), dodge=True, linewidth=1, ax=ax[1,2], color='black', jitter=False, orient='h')
sns.violinplot(data=df.loc[6]['82B11'].split(','), ax=ax[1,2], color='orange', orient='h')
ax[0,0].set_xlim(0,200)
ax[0,1].set_xlim(0,200)
ax[0,2].set_xlim(0,200)
ax[1,0].set_xlim(0,200)
ax[1,1].set_xlim(0,200)
ax[1,2].set_xlim(0,200)
ax[1,0].set_xlabel('92A11')
ax[1,1].set_xlabel('92A11')
ax[1,2].set_xlabel('92A11')
ax[0,0].set_ylabel('P8')
ax[1,0].set_ylabel('P7')
fig.tight_layout()
Seaborn works easiest with its data in long form. pd.melt can convert the 3 individual columns into a variable-name and value column. pd.explode can convert lists in cells into multiple rows. As the current entries are strings, they can be converted to lists of numbers before calling .explode().
The following code supposes you want box plots, not bars. It also supposes that an entry with only '0' is supposed to stay empty.
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
df = pd.read_csv('raw_data.csv')
df = df.rename(columns={df.columns[0]: 'Location'}) # suppose the first column is named 'Location'
df_melted = df.melt(id_vars='Location', var_name='Dataset', value_name='Datapoint')
df_melted['Datapoint'] = df_melted['Datapoint'].map(
lambda s: [np.nan] if s == '0' else [float(v) for v in s.split(',')])
df_exploded = df_melted.explode(column='Datapoint', ignore_index=True)
df_exploded['Datapoint'] = pd.to_numeric(df_exploded['Datapoint']) # make sure the column is numeric
g = sns.catplot(data=df_exploded, col='Dataset', kind='box', x='Datapoint', y='Location')
g.set(xlabel='', ylabel='') # remove x and y label
PS: df_merged looks like:
Location Dataset Datapoint
0 P1 92A11 [nan]
1 P2 92A11 [10.0, 15.0]
2 P3 92A11 [100.0, 40.0, 60.0, 50.0, 70.0, 50.0, 30.0, 40...
3 P4 92A11 [55.0, 40.0, 20.0, 15.0, 15.0, 15.0, 15.0, 10....
4 P5 92A11 [nan]
...
df_exploded has each value inside the list into a separate row:
Location Dataset Datapoint
0 P1 92A11 NaN
1 P2 92A11 10.0
1 P2 92A11 15.0
2 P3 92A11 100.0
2 P3 92A11 40.0
...