Relationship analysis issue

[Rohit_1990]

Hi All,

I am trying to create a dataset from a file such that if party has already in cluster it can not be a part of any other cluster.

Party1 party2

1 2

1 7

2 1

2 8

3 5

3 8

7 3

8 7

So basically we will start by smallest party1 and all corresponding party2 for that party1 are part of cluster.

But for party1= 1 , 2 and 7 are already part of cluster hence all cases with party1  syartist with 2and 7 will be eliminated and smallest next party 1will be considered for cluster .so next cluster will be 3 and so on

[mkeintz]

You've shown us the input data, and made a description of the requirements.

Help us help you.  Show how you want the result data set or table to appear.

[PGStats]

Does your example data implies that 1, 2, 3, 5, 7, 8 are all part of the same cluster?

[Rohit_1990]

Output data set would be
Party1 party 2
1 2
1 7
3 5
3 8

[Ksharp · Accepted solution]

If I understood your question.

data have;
input Party1 party2;
cards;
1 2
1 7
2 1
2 8
3 5
3 8
7 3
8 7
;
run;

data want;
 if _n_=1 then do;
  if 0 then set have;
  declare hash h();
  h.definekey('party2');
  h.definedone();
 end;
set have;
if h.check(key:party1)=0 then delete;
h.replace();
run;

proc print;run;

[Rohit_1990]

Hi Ksharp ,

your response is perfect as per my concern but I missed to highlight one scenario in my query.

data have;

input Party1 party2;

cards;

1 2

1 7

2 1

2 8

2 9

3 5

3 8

7 3

8 7

9 1

;

run;

I have included one more row in cluster for prty1=2 i.e. 2 9

Actually I want to excude all rows starting with prty2 values for my valid cluster.

since cluster starting with part1 =1 is valid cluster hence
all rows satrting with 2 and 7 will be discarded.

But since cluster starting with 2 was never considered to be a valid cluster
so prty2 values for this cluster rows will not be considered for row elimnation

so my final op will be something like

Party1 party2;

1 2

1 7

3 5

3 8

9 1

Thanks a ton in advance!!!!!!!!!!!!!!!!!!!!!!!

[Ksharp]

OK. How about this one ?

data have;
input Party1 party2;
cards;
1 2
1 7
2 1
2 8
2 9
3 5
3 8
7 3
8 7
9 1
;
run;

 
data want;
 if _n_=1 then do;
  if 0 then set have;
  declare hash h();
  h.definekey('party2');
  h.definedone();
 end;
set have;
if h.check(key:party1)=0 then delete;
 else h.replace();
run;

proc print;run;