normalize data and plot as stacked bar plot with python/ pandas

Question

I have the following data frame:

    land_cover  canopy_cat  count  tc_density_cor
0           20           1     56       35.760967
1           20           2     28       35.760967
2           20           3     11       35.760967
3           20           4      9       35.760967
4           20           5      4       35.760967
5           20           6      3       35.760967
6           20           7      3       35.760967
7           20           8      1       35.760967
8           20           9      4       35.760967
9           20          10      6       35.760967
10          20          11      2       35.760967
11          30           1    194       17.408260
12          30           2     86       17.408260
13          30           3     55       17.408260
14          30           4     36       17.408260
15          30           5     21       17.408260
16          30           6     15       17.408260
17          30           7      9       17.408260
18          30           8      6       17.408260
19          30           9     19       17.408260
20          30          10     14       17.408260
21          30          11      9       17.408260
22          40           1    106       17.458283
23          40           2     45       17.458283
24          40           3     19       17.458283
25          40           4     14       17.458283
26          40           5      9       17.458283
27          40           6      8       17.458283
28          40           7      5       17.458283
29          40           8      5       17.458283
30          40           9      8       17.458283
31          40          10     12       17.458283
32          40          11      3       17.458283

and I want to plot my data as a stacked bar plot:
x-axis = land_cover
y-axis = count per canopy_cat

I think that the pivot function is what I am looking for. However before I want to normalize the "count" column for each land_cover relative to "tc_density_cor".
for example, the sum of "counts" for land_cover=20 = 127.
127/35.76 = 56/x --> new value would be: 15.76

How can I do that?? :)

Answer 1

IIUC

d = df.set_index(
    ['land_cover', 'canopy_cat']
).pipe(
    lambda d: d['count'].div(d['count'].sum(level=0), axis=0, level=0) * d['tc_density_cor']
).unstack()

d.iloc[:, :5]

canopy_cat          1         2         3         4         5
land_cover                                                   
20          15.768615  7.884308  3.097407  2.534242  1.126330
30           7.278454  3.226531  2.063479  1.350641  0.787874
40           7.908453  3.357362  1.417553  1.044513  0.671472

---

d.plot.bar(stacked=True)

---

Same answer refactored

def normalize(d):
    sums = d['count'].sum(level='land_cover')
    return d['count'].div(sums, axis=0, level='land_cover') * d['tc_density_cor']

d = df.set_index(['land_cover', 'canopy_cat']).pipe(normalize).unstack()

d.plot.bar(stacked=True)

Answer 2

I think you need:

df['Count Per Canopy Cat'] = (df['count'] * df['tc_density_cor'] / 
                              df.groupby('land_cover')['count'].transform(sum))

df.pivot('land_cover',
         'canopy_cat',
         'Count Per Canopy Cat')\
  .plot.bar(stacked=True, figsize=(15,8))

Chart: