Skip to content

Commit

Permalink
Merge pull request #337 from muridaehsr/RoundingHackathon
Browse files Browse the repository at this point in the history 
Rounding hackathon
  • Loading branch information
@statasaurus
statasaurus authored Oct 21, 2024
2 parents 9de996a + a3e5a42 commit 578d538
Show file tree
Hide file tree
Showing 2 changed files with 38 additions and 23 deletions.
5 changes: 2 additions & 3 deletions Comp/r-sas_rounding.qmd
View file
Original file line number Diff line number Diff line change
Expand Up @@ -4,10 +4,9 @@ title: "R v SAS rounding"

**Rounding; R and SAS**

On comparing the documentation of rounding rules for both languages, it will be noted that the default rounding rule (implemented in the respective language's round() function) are different. Numerical differences arise in the knife-edge case where the number being rounded is equidistant between the two possible results. The round() function in SAS will round the number 'away from zero', meaning that 12.5 rounds to the integer 13. The round() function in Base R will round the number 'to even', meaning that 12.5 rounds to the integer 12. SAS does provide the rounde() function which rounds to even and the janitor package in R contains a function round_half_up() that rounds away from zero. In this use case, SAS produces a correct result from its round() function, based on its documentation, as does R. Both are right based on what they say they do, but they produce different results (Rimler, M.S. et al.).

As described in [R - Rounding](../R/rounding.html) and [SAS - Rounding](../SAS/rounding.html), `round()` in SAS and `round_half_up()` in R are incorrect in rare different cases. It may need to be considered when comparing results between SAS and R.
On comparing the documentation of rounding rules for both languages, it will be noted that the default rounding rule (implemented in the respective language's round() function) are different. Numerical differences arise in the knife-edge case where the number being rounded is equidistant between the two possible results. The round() function in SAS will round the number 'away from zero', meaning that 12.5 rounds to the integer 13. The round() function in Base R will round the number 'to even', meaning that 12.5 rounds to the integer 12. SAS does provide the rounde() function which rounds to even and the cards package in R contains a function round5() that rounds away from zero. In this use case, SAS produces a correct result from its round() function, based on its documentation, as does R. Both are right based on what they say they do, but they produce different results (Rimler, M.S. et al.).

As described in [R - Rounding](../R/rounding.html) and [SAS - Rounding](../SAS/rounding.html), `round()` in SAS and `cards::round5()` in R are incorrect in rare different cases. It may need to be considered when comparing results between SAS and R.

**References**

Expand Down
56 changes: 36 additions & 20 deletions SAS/rounding.qmd
View file
Original file line number Diff line number Diff line change
Expand Up @@ -15,32 +15,47 @@ For example (See references for source of the example)
```{r, eval=FALSE}
#Example code
data XXX;
my_number=2.2; output;
my_number=3.99; output;
my_number=1.2345; output;
my_number=7.876; output;
my_number=13.8739; output;
my_number=2.2; decimal_paces=1; output;
my_number=2.2; decimal_paces=2; output;
my_number=2.2;; decimal_paces=3 output;
my_number=3.99; decimal_paces=1; output;
my_number=3.99; decimal_paces=2; output;
my_number=3.99; decimal_paces=3; output;
my_number=1.2345; decimal_paces=1; output;
my_number=1.2345; decimal_paces=2; output;
my_number=1.2345; decimal_paces=3; output;
my_number=7.876; decimal_paces=1; output;
my_number=7.876; decimal_paces=2; output;
my_number=7.876; decimal_paces=3; output;
my_number=13.8739; decimal_paces=1; output;
my_number=13.8739; decimal_paces=2; output;
my_number=13.8739; decimal_paces=3; output;
run;
data xxx2;
set xxx;
r_1_dec = round(my_number, 0.1);
r_2_dec = round(my_number, 0.01);
r_3_dec = round(my_number, 0.001);
re_1_dec = rounde(my_number, 0.1);
re_2_dec = rounde(my_number, 0.01);
re_3_dec = rounde(my_number, 0.001);
round = round(my_number, decimal_places);
rounde = rounde(my_number, decimal_places);
run;
```

| my_number | r_1_dec | r_2_de | r_3_dec | re_1_dec | **re_2_dec** | **re_3_dec** |
|------------|---------|--------|-----------|----------|--------------|--------------|
| 2.2 | 2.2 | 2.2 | 2.2 | 2.2 | 2.2 | 2.2 |
| 3.99 | 4 | 3.99 | 3.99 | 4 | 3.99 | 3.99 |
| **1.2345** | 1.2 | 1.23 | **1.235** | 1.2 | 1.23 | **1.234** |
| 7.876 | 7.9 | 7.88 | 7.876 | 7.9 | 7.88 | 7.876 |
| 13.8739 | 13.9 | 13.87 | 13.874 | 13.9 | 13.87 | 13.874 |
| my_number | decimal_places | round | rounde |
|------------|----------------|-----------|-----------|
| 2.2 | 0.1 | 2.2 | 2.2 |
| 2.2 | 0.01 | 2.2 | 2.2 |
| 2.2 | 0.001 | 2.2 | 2.2 |
| 3.99 | 0.1 | 4 | 4 |
| 3.99 | 0.01 | 3.99 | 3.99 |
| 3.99 | 0.001 | 3.99 | 3.99 |
| **1.2345** | 0.1 | 1.2 | 1.2 |
| **1.2345** | 0.01 | 1.23 | 1.23 |
| **1.2345** | 0.001 | **1.235** | **1.234** |
| 7.876 | 0.1 | 7.9 | 7.9 |
| 7.876 | 0.01 | 7.88 | 7.89 |
| 7.876 | 0.001 | 7.876 | 7.876 |
| 13.8739 | 0.1 | 13.9 | 13.9 |
| 13.8739 | 0.01 | 13.87 | 13.87 |
| 13.8739 | 0.001 | 13.874 | 13.874 |

In some rare cases, `round()` does not return result as expected. For example below.

Expand All @@ -54,6 +69,7 @@ run;
```

You can find a little more by the code below. It creates dummy numbers with different numbers of decimal digits, and filter incorrect results. Note, the incorrect results are expected when the input number is near or beyond the precision level, i.e. the last decimal of the input number is near or less than the number multiplied by `constant('maceps')`.

```{sas, eval=FALSE}
data dum1;
int1=0; output;
Expand All @@ -80,6 +96,7 @@ quit;
```

Or more by the code below and comparing with results from another language, e.g. R.

```{sas, eval=FALSE}
data dum1;
dec1=0; int1=0; output;
Expand Down Expand Up @@ -113,7 +130,6 @@ data dat3;
run;
```


As the incorrect rounding all occur on large number, `round()` is still reliable in most of cases.

**References**
Expand Down

0 comments on commit 578d538

Please sign in to comment.