使用seaborn为数据帧绘制直方图

Question

I have a dataFrame which has multiple columns and many rows..Many row has no value for column so in the data frame its represented as NaN. The example dataFrame is as follows,

df.head()
GEN Sample_1    Sample_2    Sample_3    Sample_4    Sample_5    Sample_6    Sample_7    Sample_8    Sample_9    Sample_10   Sample_11   Sample_12   Sample_13   Sample_14
A123    9.4697  3.19689 4.8946  8.54594 13.2568 4.93848 3.16809 NAN NAN NAN NAN NAN NAN NAN
A124    6.02592 4.0663  3.9218  2.66058 4.38232         NAN NAN NAN NAN NAN NAN NAN
A125    7.88999 2.51576 4.97483 5.8901  21.1346 5.06414 15.3094 2.68169 8.12449 NAN NAN NAN NAN NAN
A126    5.99825 10.2186 15.2986 7.53729 4.34196 8.75048 16.9358 5.52708 NAN NAN NAN NAN NAN NAN
A127    28.5014 4.86702 NAN NAN NAN NAN NAN NAN NAN NAN NAN NAN NAN NAN

I wanted to plot histogram for this dataFrame using seaborn function from python and so i was trying the following lines,

sns.set(color_codes=True)
sns.set(style="white", palette="muted")
sns.distplot(df)

But its throwing the following error,

    ValueError                                Traceback (most recent call last)
    <ipython-input-80-896d7fe85ef3> in <module>()
          1 sns.set(color_codes=True)
          2 sns.set(style="white", palette="muted")
    ----> 3 sns.distplot(df)

    /anaconda3/lib/python3.4/site-packages/seaborn/distributions.py in distplot(a, bins, hist, kde, rug, fit, hist_kws, kde_kws, rug_kws, fit_kws, color, vertical, norm_hist, axlabel, label, ax)
        210         hist_color = hist_kws.pop("color", color)
        211         ax.hist(a, bins, orientation=orientation,
    --> 212                 color=hist_color, **hist_kws)
        213         if hist_color != color:
        214             hist_kws["color"] = hist_color

   /anaconda3/lib/python3.4/site-packages/matplotlib/axes/_axes.py in hist(self, x, bins, range, normed, weights, cumulative, bottom, histtype, align, orientation, rwidth, log, color, label, stacked, **kwargs)
       5627             color = mcolors.colorConverter.to_rgba_array(color)
       5628             if len(color) != nx:
    -> 5629                 raise ValueError("color kwarg must have one color per dataset")
       5630 
       5631         # We need to do to 'weights' what was done to 'x'

    ValueError: color kwarg must have one color per dataset

Any helps/suggestions to get rid of this error would be greatly appreciated..!!!

Answer 1

Let's assume I have the excerpt from the data you have showed above (with only difference that on my machine NAN is NaN ).

Then, the best graphical representation I can think of is grouped barplot: one group for every sample, within every group there are gene bars (some people call this histogram occasionally)

In order to do that, you need first to "melt" your data, in R parlour, ie make it "long". Then, you can proceed with plotting.

data = df.set_index('GEN').unstack().reset_index()
data.columns = ['sample','GEN', 'value']

sns.set(style="white")
g = sns.factorplot(x='sample'
                   ,y= 'value'
                   ,hue='GEN'
                   ,data=data
                   ,kind='bar'
                   ,aspect=2
                   )
g.set_xticklabels(rotation=30);

Please, let us know if this is the type of plot you were after.

Answer 2

I had also thought the seaborn documentation mentioned that multiple columns could be plotted simultaneously, and highlighted by color by default.

But upon re-reading, I did not see anything. Instead, I think I inferred it from this tutorial , where part of the way through, the tutorial plots a data frame with multiple columns.

---

However, the "solution" is trivial, and hopefully exactly what you're looking for:

sns.set(color_codes=True)
sns.set(style="white", palette="muted")
sns.distplot(df)

for col_id in df.columns:
    sns.distplot(df[col_id])

By default, this will alter the colors, "knowing" which one has already been used.

Note: I used a different data set, since I wasn't sure how to re-create yours.

Answer 3

I had similar problem because my pandas.DataFrame had elements of type Object in a column I wanted to plot ( my_column ). So that the command:

print(df[my_column])

gave me:

Length: 150, dtype: object

The solution was

sns.distplot(df[my_column].astype(float))

As the datatype of my_column transformed to:

Length: 150, dtype: float64