Controlled rounding
Rounding is performed independently on each cell (including subtotals or grand total). See Step 3 below. Here is a simple example of how rounding is performed:
Suppose we have a survey of 10 people representing a population of 1,205 and let 100 be the rounding base used by the RTRA.
Microdata
Microdata
Unit | Sex | Marital status | Weight |
---|---|---|---|
1 | Male | Married | 120.5 |
2 | Male | Widowed | 120.5 |
3 | Male | Married | 120.5 |
4 | Male | Divorced | 120.5 |
5 | Male | Married | 120.5 |
6 | Female | Married | 120.5 |
7 | Female | Married | 120.5 |
8 | Female | Widowed | 120.5 |
9 | Female | Widowed | 120.5 |
10 | Female | Married | 120.5 |
Step 1
Step 1
The RTRA modified SAS procedure creates the following table:
Married | Divorced | Widowed | Total | |
---|---|---|---|---|
Male | 361.5 | 120.5 | 120.5 | 602.5 |
Female | 361.5 | 0.0 | 241.0 | 602.5 |
Total | 723.0 | 120.5 | 361.5 | 1,205.0 |
Step 2
Step 2
The RTRA system applies conventional rounding to the inner cells, and then sums up totals and subtotals.
Married | Divorced | Widowed | Total | |
---|---|---|---|---|
Male | 362 | 121 | 121 | 603 |
Female | 362 | 0 | 241 | 603 |
Total | 724 | 121 | 362 | 1,207 |
Step 3
Step 3
The RTRA system sends these counts into the controlled random rounding program. We obtain the following:
Married | Divorced | Widowed | Total | |
---|---|---|---|---|
1.The order of specification of variables in the procedure may have an effect on the rounded results. For instance, specifying sex and marital status might give different results from those obtained by specifying marital status and sex. In both cases rounding is controlled. | ||||
Male | 400 | 100 | 100 | 600 |
Female | 300 | 0 | 300 | 600 |
Total | 700 | 100 | 400 | 1,200 |
Note that rows and columns are additives. This last table is then sent to the researcher.
- Date modified: