drop table tbl_family;

create table tbl_family
(
father nvarchar2(50) ,
mother nvarchar2(50) ,
child nvarchar2(50) ,
birth date null
);

insert into tbl_family (father,mother,child,birth) values ('Ralf', 'Nina', 'Adam', to_date( '17-02-1989' , 'dd-MM-yyyy') );
insert into tbl_family (father,mother,child,birth) values ('Ralf', 'Nina', 'Emma', null) ;
insert into tbl_family (father,mother,child,birth) values ('Ralf', 'Birgit', 'Andi', to_date( '09-06-1982' , 'dd-MM-yyyy') );
insert into tbl_family (father,mother,child,birth) values ('Alec', 'Birgit', 'Amos', to_date( '18-05-1977',  'dd-MM-yyyy')) ;
insert into tbl_family (father,mother,child,birth) values ('Aaron', 'Nanci', 'Alfred', to_date( '28-10-1934',  'dd-MM-yyyy') );
insert into tbl_family (father,mother,child,birth) values ('Aaron', 'Rike', 'Ashley', null) ;
insert into tbl_family (father,mother,child,birth) values ('Andy', 'Celine', 'Roland', to_date( '20-12-1952',  'dd-MM-yyyy') );

Adam/Emma/Andi 
Andi/Amos
Alfred/Ashley

I'm not sure why you want to use connect by? You've only got one level in the hierarchy!

It seems like listagg is more what you need. You can use grouping sets to get the list of children for each mother and father in one go:

select coalesce ( mother, father ) par, 
       listagg(child, ',') within group (order by child) sibs
from   tbl_family
group  by grouping sets ( (mother), (father) )
having count(*) > 1;

PAR      SIBS                         
Birgit   Amos,Andi
Nina     Adam,Emma
Aaron    Alfred,Ashley
Ralf     Adam,Andi,Emma

But this includes children for one parent which a are subset of another's (Nina and Ralf's kids).

If you want to avoid this, you need to exclude children who have more siblings from one parent than the other.

One way to do this to:

- Use the collect function to load the children into an array
- Use submultiset to exclude siblings that are a subset of another

Sadly, collect doesn't support grouping sets yet (run on 12.2.0.1):

create or replace type arr is table of varchar2(50);
/

select mother, father, 
       cast ( collect ( cast ( child as varchar2(50) ) ) as arr ) c,
       listagg(child, ',') within group (order by child) 
from   tbl_family
group  by grouping sets ( mother , father );

SQL Error: ORA-03001: unimplemented feature

You can workaround this by collecting the mothers and fathers separately. Then unioning them back together.

You then return those siblings that aren't a submultiset of another:

with fathers as ( 
  select father par, 
         cast ( collect ( cast ( child as varchar2(50) ) ) as arr ) c,
         listagg(child, ',') within group (order by child) sibs
  from   tbl_family
  group  by father
  having count(*) > 1
), mothers as ( 
  select mother par, 
         cast ( collect ( cast ( child as varchar2(50) ) ) as arr ) c,
         listagg(child, ',') within group (order by child) sibs
  from   tbl_family
  group  by mother
  having count(*) > 1
), children as (
  select * from fathers
  union all 
  select * from mothers
)
  select distinct sibs from children c
  where  not exists (
    select * from children subc
    where  c.c submultiset of subc.c 
    and    c.sibs <> subc.sibs
  );

SIBS                         
Alfred,Ashley
Adam,Andi,Emma
Amos,Andi

Obviously, as you're working with given names it's easy to have siblings that are a subset of an unrelated group of siblings... Even when you throw in their family names and their parent's you can end up with spurious subsets. Beware!