Python: plot each element in lists which are in a dataframe with seaborn facetWrap

Question

I have a dataframe containing lists of elements like this:

I have 10 stimulus values and 16 trials per stimulus. The lists contain spike amplitude values of a spike train. So if there are e.g. two values 33 and 34 in a list, my algorithm found two spikes with 33mV and 34mV. I want to plot all of those values to get an impression of my data.

Currently I'm using these lines of code to get this plot

flatui = ["#9b59b6", "#3498db", "#95a5a6", "#e74c3c", "#34495e", "#2ecc71"]
g = sns.FacetGrid(df, col='stimulus', col_wrap=5, sharey=True, sharex=True)
g.map(sns.swarmplot, 'trial', 'data', edgecolor="black", linewidth=.9, palette=flatui)
g.set_axis_labels('trials', '[ms]')

Obviously it doesn't plot all values but only one per list (the mean? the first? I don't know). Can you help me to plot all of them at once?

Build my dataframe

trial_vec       = np.tile(np.arange(16)+1, 10)     
stimulus_vec    = np.repeat([-2., -1.75, -1., -0.75, -0.5,  0.5,  1.,  1.25,  1.75,  2.5 ], 16)                  
data_vec        = np.random.randint(0, 16, size=160)
spi_amp         = pd.DataFrame({'trial': trial_vec, 'stimulus': stimulus_vec, 'data': data_vec}).astype('object')
spi_amp["data"] = [np.random.rand(4).tolist() for i in range(160)]
spi_amp

In your last questions you did provide a [mcve]. Why not here? Without it, it's really hard to even understand the issue. — ImportanceOfBeingErnest, Sep 29 '17 at 12:37
I agree. The reason is I don't know how to put a two dimensional list into a dataframe to that i can replicate the datarame I have. I created the data vector like this: np.random.random((160, 4)) but panda doesn't accept 2-D arrays just like that. — Svenno Nito, Sep 29 '17 at 12:59
Try something like `df = pd.DataFrame({"x" : np.linspace(0,1,50)}); df["listcol"] = [np.random.rand(4).tolist() for i in range(50)]`. — ImportanceOfBeingErnest, Sep 29 '17 at 13:03

score 2 · Accepted Answer · answered Sep 29 '17 at 20:54

Thinking about it, I guess the only way to accomplish what you want is to convert the dataframe with the lists in one column into a long form, such that each list entry has its own row in the dataframe. See this question on a possible way of doing this.

Using the presented method on the data from the question gives you:

import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

trial_vec       = np.tile(np.arange(16)+1, 10)     
stimulus_vec    = np.repeat([-2., -1.75, -1., -0.75, -0.5,  0.5,  1.,  1.25,  1.75,  2.5 ], 16)                  
data_vec        = np.random.randint(0, 16, size=160)
df         = pd.DataFrame({'trial': trial_vec, 'stimulus': stimulus_vec}).astype('object')
df["data"] = [np.random.rand(4).tolist() for i in range(160)]

# convert dataframe to long form
df2 = df.set_index(['trial', 'stimulus'])['data'].apply(pd.Series).stack()
df2 = df2.reset_index()
df2.columns = ['trial', 'stimulus','pos','data']

# the plotting code stays the same
flatui = ["#9b59b6", "#3498db", "#95a5a6", "#e74c3c", "#34495e", "#2ecc71"]
g = sns.FacetGrid(df2, col='stimulus', col_wrap=5, sharey=True, sharex=True)
g.map(sns.swarmplot, 'trial', 'data', edgecolor="black", linewidth=.9, palette=flatui)
g.set_axis_labels('trials', '[ms]')

plt.show()

Couldn't have asked for more, perfect result. I owe you. – Svenno Nito Sep 30 '17 at 11:58 — Svenno Nito, Sep 30 '17 at 11:58

Python: plot each element in lists which are in a dataframe with seaborn facetWrap

Build my dataframe

1 Answers1