pandas: Filter a whole group out based on a condition

I've created the following MWE

data = {'Name': ['Tom', 'Tom', 'Tom', 'Tom', 'Tom', 'Tom', 'Tom' , 'Tom', 'Tom', 'Tom'], 'Article': ['A', 'A', 'A', 'B', 'B', 'A', 'A', 'A', 'B', 'B'], 'Weekday' : [1,2,3,2,3,1,2,3, 1, 2], 'Value': [1,40,3,91,10,6,9,10, 20, 10]}  

df_test = pd.DataFrame(data)

	    Name	Article	Weekday	Value
    0	Tom	       A	1	    1
    1	Tom	       A	2	   40
    2	Tom	       A    3	    3
    3	Tom        B	2	   91
    4	Tom	       B	3	   10
    5	Tom	       A	1	    6
    6	Tom	       A	2	    9
    7	Tom	       A	3	   10
    8	Tom	       B	1	   20
    9	Tom	       B	2	   10

where a group consists of Name-Article pairs. I want to filter all groups out that dont have at least 2 counts in values for all three existing weekdays. So there should be only groups having three weekdays (1,2,3) and with at least two counts. If a Name-Article has only two weekdays with at least two counts it should be filtered out as well.

The expected output should look like this

	    Name	Article	Weekday	Value
    0	Tom	       A	1	    1
    1	Tom	       A	2	   40
    2	Tom	       A    3	    3
    5	Tom	       A	1	    6
    6	Tom	       A	2	    9
    7	Tom	       A	3	   10

If you want to ensure at least 2 counts per Name/Article per Weekday, you can compute a crosstab to count the combinations of Name/Article and Weekday.

Then you can use any filter you want, here we keep the Name/Article combinations that have all at least 2 values:

counts = pd.crosstab([df_test['Name'], df_test['Article']], df_test['Weekday'])
keep = counts[counts.ge(2).all(axis=1)]

out = df_test.set_index(['Name', 'Article']).loc[keep.index].reset_index()

# or
# out = df_test.merge(keep[[]].reset_index())

To count not all days but only a given number (e.g. ≥3), use:

keep = counts[counts.ge(2).sum(axis=1).ge(3)]

Output:

  Name Article  Weekday  Value
0  Tom       A        1      1
1  Tom       A        2     40
2  Tom       A        3      3
3  Tom       A        1      6
4  Tom       A        2      9
5  Tom       A        3     10

Intermediate counts:

Weekday       1  2  3
Name Article         
Tom  A        2  2  2  # all 3 have ≥ 2, we keep
     B        1  2  1  # not all ≥ 2, discard

Use boolean indexing for filtering:

s = df_test.groupby(['Name','Article','Weekday']).size()
m = s.ge(2).groupby(level=[0,1]).sum().ge(3)
df = df_test[df_test.set_index(['Name', 'Article']).index.isin(m.index[m])]
print (df)
  Name Article  Weekday  Value
0  Tom       A        1      1
1  Tom       A        2     40
2  Tom       A        3      3
5  Tom       A        1      6
6  Tom       A        2      9
7  Tom       A        3     10

How it working:

#get counts per Name/Article/Weekday
print (df_test.groupby(['Name','Article','Weekday']).size())
Name  Article  Weekday
Tom   A        1          2
               2          2
               3          2
      B        1          1
               2          2
               3          1
dtype: int64

#test if greater or equal like 2
print (s.ge(2))
Tom   A        1           True
               2           True
               3           True
      B        1          False
               2           True
               3          False
dtype: bool

#count Trues per Name/Article
print (s.ge(2).groupby(level=[0,1]).sum())
Tom   A          3
      B          1
dtype: int64

#test if counts greater/ equal like 3
print (s.ge(2).groupby(level=[0,1]).sum().ge(3))
Tom   A           True
      B          False
      
#filter Name/Article
print (m.index[m])
MultiIndex([('Tom', 'A')],
           names=['Name', 'Article'])

#filter combinaton from original DataFrame Name/Article
print (df_test.set_index(['Name', 'Article']).index.isin(m.index[m]))
[ True  True  True False False  True  True  True False False]

Similar solution if need test if all NAme/Article has 2 or more counts with GroupBy.all:

s = df_test.groupby(['Name','Article','Weekday']).size()
m = s.ge(2).groupby(level=[0,1]).all()

df1 = df_test[df_test.set_index(['Name', 'Article']).index.isin(m.index[m])]
print (df1)
  Name Article  Weekday  Value
0  Tom       A        1      1
1  Tom       A        2     40
2  Tom       A        3      3
5  Tom       A        1      6
6  Tom       A        2      9
7  Tom       A        3     10