[FIXED] Matplotlib grouped bar chart with individual data points

Issue

I am trying to show individual data points in a grouped bar chart using Matplotlib. I tried to do it with a scatterplot and I found a related stackoverflow topic: pyplot bar charts with individual data points. However it only provides a solution for regular bar charts, but not for grouped bar charts.

This is my code generating a grouped bar chart without error bars:

# Create a list for on_target, ntc, on_target_error, and ntc_error
on_target = [df_subset['primer_pair_1_on_target'][36], df_subset['primer_pair_2_on_target'][36], df_subset['primer_pair_3_on_target'][36], df_subset['primer_pair_4_on_target'][36], df_subset['primer_pair_5_on_target'][36], df_subset['primer_pair_6_on_target'][36], df_subset['primer_pair_7_on_target'][36], df_subset['primer_pair_8_on_target'][36], df_subset['primer_pair_9_on_target'][36]]
ntc = [df_subset['primer_pair_1_NTC'][36], df_subset['primer_pair_2_NTC'][36], df_subset['primer_pair_3_NTC'][36], df_subset['primer_pair_4_NTC'][36], df_subset['primer_pair_5_NTC'][36], df_subset['primer_pair_6_NTC'][36], df_subset['primer_pair_7_NTC'][36], df_subset['primer_pair_8_NTC'][36], df_subset['primer_pair_9_NTC'][36]]
on_target_error = [df_subset['primer_pair_1_on_target_error'][36], df_subset['primer_pair_2_on_target_error'][36], df_subset['primer_pair_3_on_target_error'][36], df_subset['primer_pair_4_on_target_error'][36], df_subset['primer_pair_5_on_target_error'][36], df_subset['primer_pair_6_on_target_error'][36], df_subset['primer_pair_7_on_target_error'][36], df_subset['primer_pair_8_on_target_error'][36], df_subset['primer_pair_9_on_target_error'][36]]
ntc_error = [df_subset['primer_pair_1_NTC_error'][36], df_subset['primer_pair_2_NTC_error'][36], df_subset['primer_pair_3_NTC_error'][36], df_subset['primer_pair_4_NTC_error'][36], df_subset['primer_pair_5_NTC_error'][36], df_subset['primer_pair_6_NTC_error'][36], df_subset['primer_pair_7_NTC_error'][36], df_subset['primer_pair_8_NTC_error'][36], df_subset['primer_pair_9_NTC_error'][36]]

# Create a variable with the x locations for the primer pairs
index1 = ["F1R1", "F2R1", "F3R1", "F1R2", "F2R2", "F3R2", "F1R3", "F2R3", "F3R3"]
ind = np.arange(len(on_target))

# Style, axis and title
plt.style.use('classic')
fig, axis = plt.subplots()
plt.ylabel("RFUs")
axes = plt.gca()
axes.set_ylim([0,3000000])
axis.set_xticks(ind)
axis.set_xticklabels(index1)
axis.yaxis.grid(True)
axis.set_axisbelow(True)
plt.title('Primer Screen')

# Layout, and bar width
fig.tight_layout()
width = 0.35  # the width of the bars

# Create on_target and ntc_bar
on_target_bar = axis.bar(ind - width/2, on_target, width, yerr=on_target_error,
            label='on-target', color='red')
ntc_bar = axis.bar(ind + width/2, ntc, width, yerr=ntc_error,
            label='ntc', color="grey")

# Create legend
axis.legend()

# Add scientific notation
mf = mpl.ticker.ScalarFormatter(useMathText=True)
mf.set_powerlimits((-2,2))
plt.gca().yaxis.set_major_formatter(mf)

plt.show()

Right now, my grouped bar chart looks like this: Primer Screen. However, I would like to include individual data points.

Thank you for any advice!

Solution

I'm not sure how you are generating your bar chart, but if you are okay with using seaborn (seaborn) you can combine the sns.barplot with the sns.stripplot like so:

import seaborn as sns

# Load some example data
tips = sns.load_dataset("tips")

Plot the chart:

# Draw the bar chart
ax = sns.barplot(
    data=tips, 
    x="day", 
    y="total_bill", 
    hue="sex", 
    alpha=0.7, 
    ci=None,
)

# Get the legend from just the bar chart
handles, labels = ax.get_legend_handles_labels()

# Draw the stripplot
sns.stripplot(
    data=tips, 
    x="day", 
    y="total_bill", 
    hue="sex", 
    dodge=True, 
    edgecolor="black", 
    linewidth=.75,
    ax=ax,
)
# Remove the old legend
ax.legend_.remove()
# Add just the bar chart legend back
ax.legend(
    handles,
    labels,
    loc=7,
    bbox_to_anchor=(1.25, .5),
)

Which produces:

By default the bar chart is plotting the mean of the data.

Answered By - Alex