Create Contour Plot from Pandas Groupby Dataframe

问题

I have following Pandas Dataframe:

In [66]: hdf.size()
Out[66]:
a           b
0           0.0          21004
            0.1         119903
            0.2         186579
            0.3         417349
            0.4         202723
            0.5         100906
            0.6          56386
            0.7           6080
            0.8           3596
            0.9           2391
            1.0           1963
            1.1           1730
            1.2           1663
            1.3           1614
            1.4           1309
...
186         0.2         15
            0.3          9
            0.4         21
            0.5          4
187         0.2          3
            0.3         10
            0.4         22
            0.5         10
188         0.0         11
            0.1         19
            0.2         20
            0.3         13
            0.4          7
            0.5          5
            0.6          1
Length: 4572, dtype: int64

You see, a from 0...188 and b in every group from some value to some value. And as the designated Z-value, the count of the occurence of the pair a/b.

How to get a countour or heatmap plot out of the grouped dataframe?

I have this (asking for the ?):

numcols, numrows = 30, 30
xi = np.linspace(0, 200, numcols)
yi = np.linspace(0, 6, numrows)
xi, yi = np.meshgrid(xi, yi)
zi = griddata(?, ?, hdf.size().values, xi, yi)

How to get the x and y values out of the Groupby object and plot a contour?

回答1:

Thanks a lot! My fault was, that I did not realize, that I have to apply some function to the groupby dataframe, like .size(), to work with it...

hdf = aggdf.groupby(['a','b']).size()
hdf

gives me

a           b
1           -2.0          1
            -1.9          1
            -1.8          1
            -1.7          2
            -1.6          5
            -1.5         10
            -1.4          9
            -1.3         21
            -1.2         34
            -1.1         67
            -1.0         65
            -0.9         94
            -0.8        180
            -0.7        242
            -0.6        239
...
187          0.4        22
             0.5        10
188         -0.6         2
            -0.5         2
            -0.4         1
            -0.3         2
            -0.2         5
            -0.1        10
            -0.0        18
             0.1        19
             0.2        20
             0.3        13
             0.4         7
             0.5         5
             0.6         1
Length: 8844, dtype: int64

With that, and your help CT Zhu, I could then do

hdfreset = hdf.reset_index()
hdfreset.columns = ['a', 'b', 'occurrence']
hdfpivot=hdfreset.pivot('a', 'b')

and this finally gave me the correct values to

X=hdfpivot.columns.levels[1].values
Y=hdfpivot.index.values
Z=hdfpivot.values
Xi,Yi = np.meshgrid(X, Y)
plt.contourf(Yi, Xi, Z, alpha=0.7, cmap=plt.cm.jet);

which leads to this beautiful contourf:

回答2:

Welcome to SO.

It looks quite clear that for each of your 'a' level, the numbers of 'b' levels are not the same, thus I will suggest the following solution:

In [44]:

print df #an example, you can get your dataframe in to this by rest_index()
    a  b     value
0   0  1  0.336885
1   0  2  0.276750
2   0  3  0.796488
3   1  1  0.156050
4   1  2  0.401942
5   1  3  0.252651
6   2  1  0.861911
7   2  2  0.914803
8   2  3  0.869331
9   3  1  0.284757
10  3  2  0.488330

[11 rows x 3 columns]
In [45]:
#notice that you will have some 'NAN' values
df=df.pivot('a', 'b', 'value')
In [46]:

X=df.columns.values
Y=df.index.values
Z=df.values
x,y=np.meshgrid(X, Y)
plt.contourf(x, y, Z) #the NAN will be plotted as white spaces
Out[46]:
<matplotlib.contour.QuadContourSet instance at 0x1081385a8>

来源：https://stackoverflow.com/questions/24032282/create-contour-plot-from-pandas-groupby-dataframe

标签

python

matplotlib

pandas

group-by

contour