New Contributor
Posts: 2

# homogeneous distribution

Hi,

I need help , have to make a random homogeneous distribution of the observations on the basis of the amount of a given variable.

I have a set of records that must be assigned to specific entities based on certain percentages.

Example (basetable):

ID  COUNT AMOUNT

1   1        300

2   1        150

3   2        500

4   1        200

5   3        700

entities and percentages

ENTITY  PERCENTAGE

A       20%

B       70%

C       10%

how can I assign the correct values ​​of COUNT and AMOUNT entities A, B and C in relation to their proportion in the most consistent and fair as possible ?

you may kindly direct me some examples?

Thank you so much , Isabel

Super User
Posts: 10,205

## Re: homogeneous distribution

Your post is ambiguous . What output would you like to see ?
New Contributor
Posts: 2

## Re: homogeneous distribution

Hi,

for example:
A 20% COUNT: 1.6 AMOUNT: 370
B 70% COUNT: 5.6 AMOUNT:1295
C 10% COUNT: 0.8 AMOUNT: 185

so:
ID  COUNT AMOUNT
1   1        300  => A
2   1        150  => A
3   2        500  => B
4   1        200  => C
5   3        700  => B

A =>COUNT: 2 AMOUNT: 450
B =>COUNT: 5 AMOUNT: 1200
C =>COUNT: 1 AMOUNT: 200
Frequent Contributor
Posts: 95

## Re: homogeneous distribution

[ Edited ]

Hi Isabel.

Is there a rule that determines the assignment of data to groups A, B and C? Or do you want this to be random, ending up with the proportions indicated (20% of obervations in group A, etc.)?

If the latter, then AnnMaria might have the solution: http://www.thejuliagroup.com/blog/?p=2599

Norman.
SAS 9.4 (TS1M4) X64_7PRO WIN 6.1.7601 Service Pack 1 Workstation

Super User
Posts: 10,205

## Re: homogeneous distribution

Ou God. I understand something now. It is more like a SAS/OR problem. Is there any OBJECT function ? I believe there are lots of combination suited your requirement . Whether you want the minimize number of each GROUP or want the minimize difference of each GROUP ?
Super User
Posts: 20,716

## Re: homogeneous distribution

Proc SurveySelect is generally used for sample selection and you can specify proportions but I can't see how A/B/C tie back to your original data. Otherwise it seems a bit like three samples just stacked together.

Discussion stats
• 5 replies
• 282 views
• 0 likes
• 4 in conversation