用 python 和 Pandas 重塑多个变量

Question

我有 5 个主要变量 f1_、f2_、f3_、f4_ 和 f5_ 的调查日期，每个 f*_ 组变量最多有 10 个子组，例如：f1_1、f1_2、f1_3 ... 或 f2_1、f2_2、。 .. f2_10.

我想执行一个 pivot_longer 来重塑我的数据框以便进行分析，我是 R 用户并且这样做了，我想知道如何通过 python、pandas 实现相同的输出。

df %>% 
  # Reshape data - to long
  pivot_longer(cols = all_of(ends_with(c("1","2","3", "4" ,"5"))), names_to = c("name", "check_id"), names_pattern = "(.*)(.)") %>% 
  # Reshape data - to wide
  pivot_wider(names_from = name) %>% 
  #unnest data
  unnest() %>% 
  # remove row if it has a NA value in both column
  filter_at(.vars = vars(one_of(c("f1_", "f2_"))),~ !is.na(.)) %>%
  # Crosstab 3 way
  tabyl(check_id, f1_ ,f2_ ) %>% 
  # add total row and col
  adorn_totals(c("row", "col" ))

这是所需的输出：

$No
 check_id Person 1 Person 2 Person 3 Person 4 Total
        1        2        0        0        0     2
        2        0        1        0        0     1
        3        0        0        1        0     1
        4        1        0        0        1     2
    Total        3        1        1        1     6

$Yes
 check_id Person 1 Person 2 Person 3 Person 4 Total
        1        5        0        0        0     5
        2        0        5        0        0     5
        3        0        1        2        0     3
        4        0        0        0        1     1
    Total        5        6        2        1    14

Python 示例数据

f1_ 和 f2_1 有 5 个子组

df = pd.DataFrame(
  {
    "f1_1": ["Person 1","NA","Person 1","Person 1","Person 1","Person 1","NA","Person 1", "Person 1"],
    "f1_2": ["Person 2","NA","Person 2","Person 2","Person 2","NA","NA","Person 2","Person 2"],
    "f1_3": ["Person 3","NA","NA","Person 3","Person 2","NA","NA","Person 3","NA"],
    "f1_4": ["Person 4","NA","NA","Person 4", "NA","NA","NA","Person 1","NA"],
    "f1_5": ["NA", "NA", "NA", "NA", "NA", "NA", "NA", "NA", "NA"],
    "f2_1": ["Yes", "NA", "Yes", "No", "Yes", "No", "NA", "Yes", "Yes"],
    "f2_2": ["Yes", "NA", "Yes", "No", "Yes", "NA", "NA", "Yes", "Yes"],
    "f2_3": ["Yes", "NA", "NA", "No", "Yes", "NA", "NA", "Yes", "NA"],
    "f2_4": ["Yes", "NA", "NA", "No", "NA", "NA", "NA", "No", "NA"],
    "f2_5": ["NA", "NA", "NA", "NA", "NA", "NA", "NA", "NA", "NA"]
    
  }
)

R 样本数据

df <- tibble::tribble(
       ~f1_1,      ~f1_2,      ~f1_3,      ~f1_4, ~f1_5, ~f2_1, ~f2_2, ~f2_3, ~f2_4, ~f2_5,
  "Person 1", "Person 2", "Person 3", "Person 4",    NA, "Yes", "Yes", "Yes", "Yes",    NA,
          NA,         NA,         NA,         NA,    NA,    NA,    NA,    NA,    NA,    NA,
  "Person 1", "Person 2",         NA,         NA,    NA, "Yes", "Yes",    NA,    NA,    NA,
  "Person 1", "Person 2", "Person 3", "Person 4",    NA,  "No",  "No",  "No",  "No",    NA,
  "Person 1", "Person 2", "Person 2",         NA,    NA, "Yes", "Yes", "Yes",    NA,    NA,
  "Person 1",         NA,         NA,         NA,    NA,  "No",    NA,    NA,    NA,    NA,
          NA,         NA,         NA,         NA,    NA,    NA,    NA,    NA,    NA,    NA,
  "Person 1", "Person 2", "Person 3", "Person 1",    NA, "Yes", "Yes", "Yes",  "No",    NA,
  "Person 1", "Person 2",         NA,         NA,    NA, "Yes", "Yes",    NA,    NA,    NA
  )

Answer 1

让我们试试吧：

# convert columns to multi index:
df.columns = pd.MultiIndex.from_tuples(map(tuple,df.columns.str.split('_')))

(df.where(df.ne('NA')).stack()
  .set_index('f1', append=True)
  .groupby(level=(1,2))['f2']
  .value_counts()
  .unstack(['f1'],fill_value=0)
  .assign(total=lambda x: x.sum(1))
)

输出：

  f1   Person 1  Person 2  Person 3  Person 4  total
  f2                                                
1 No          2         0         0         0      2
  Yes         5         0         0         0      5
2 No          0         1         0         0      1
  Yes         0         5         0         0      5
3 No          0         0         1         0      1
  Yes         0         1         2         0      3
4 No          1         0         0         1      2
  Yes         0         0         0         1      1

Answer 2

你也可以使用pyjanitor中的pivot_longer函数；目前你必须从github安装最新的开发版本：

 # install latest dev version
 # pip install git+https://github.com/ericmjl/pyjanitor.git

 import janitor

temp = (
    df.pivot_longer(names_to=(".value", "fs"), names_sep="_")
    .loc[lambda df: df.ne("NA").all(1)]
    .groupby(["fs", "f2", "f1"])
    .f1.agg("count")
    .unstack(fill_value=0)
    .assign(total=lambda df: df.sum(1))
    .swaplevel("f2", "fs")
)

temp

f1      Person 1  Person 2  Person 3  Person 4  total
f2  fs                                               
No  1          2         0         0         0      2
Yes 1          5         0         0         0      5
No  2          0         1         0         0      1
Yes 2          0         5         0         0      5
No  3          0         0         1         0      1
Yes 3          0         1         2         0      3
No  4          1         0         0         1      2
Yes 4          0         0         0         1      1

然后您可以将其拆分为 yes 和 no 数据框：

yes_df = temp.loc(axis=0)["Yes", :]
yes_df = yes_df.copy()
yes_df.loc[("Yes", "Total"), :] = yes_df.sum()
yes_df

f1         Person 1  Person 2  Person 3  Person 4  total
f2  fs                                                  
Yes 1           5.0       0.0       0.0       0.0    5.0
    2           0.0       5.0       0.0       0.0    5.0
    3           0.0       1.0       2.0       0.0    3.0
    4           0.0       0.0       0.0       1.0    1.0
    Total       5.0       6.0       2.0       1.0   14.0




no_df = temp.loc(axis=0)["No", :]
no_df = no_df.copy()
no_df.loc[("No", "Total"), :] = no_df.sum()
no_df

f1        Person 1  Person 2  Person 3  Person 4  total
f2 fs                                                  
No 1           2.0       0.0       0.0       0.0    2.0
   2           0.0       1.0       0.0       0.0    1.0
   3           0.0       0.0       1.0       0.0    1.0
   4           1.0       0.0       0.0       1.0    2.0
   Total       3.0       1.0       1.0       1.0    6.0

对于您在 R 中的代码，我能否建议您对现有代码进行一些调整：

df %>% 
  pivot_longer(starts_with('f'), 
               names_to = c('.value', "check_id"), 
               names_sep="_")%>%
  drop_na()%>%
  tabyl(check_id, f1, f2)%>%
  adorn_totals(c("row", "col"))

$No
 check_id Person 1 Person 2 Person 3 Person 4 Total
        1        2        0        0        0     2
        2        0        1        0        0     1
        3        0        0        1        0     1
        4        1        0        0        1     2
    Total        3        1        1        1     6

$Yes
 check_id Person 1 Person 2 Person 3 Person 4 Total
        1        5        0        0        0     5
        2        0        5        0        0     5
        3        0        1        2        0     3
        4        0        0        0        1     1
    Total        5        6        2        1    14