BookmarkSubscribeRSS Feed
0 Likes

Hi, everyone.

 

Recently, I found that there is a translation error of the Odds Ratio Estimates results from PROC LOGISTIC in chinese version. I use the data set CARS form SASHELP library to do a demo, just ignore the model effect. Here is the SAS code:

 

data test;
    set sashelp.cars;
    where DriveTrain in ('Front', 'Rear');
run;

proc logistic data=test plots(only)=all;
    class type(param=ref) origin(param=ref);
    model DriveTrain (event = 'Front')=msrp enginesize horsepower mpg_highway origin;
run;

 

 

The Odds Ratio Estimates from English version:

EnglishVersion.png

 

For the OR estimate of CLASS variable Origin, 'USA' is the reference level.

  • Asia vs USA
  • Europe vs USA

This is correct. But for Chinese version, here is the result:

ChineseVersion.png

 

  • USA-Asia
  • USA-Europe

 

At first I couldn't understand the result, why it's "USA-Asia"? "USA" is the reference level, it makes me confused. Somehow I find it is correct in English version.

 

So, I guess this is a translation error. Or is there some other reason?

9 Comments
ballardw
Super User

Can you please provide more details of what you think is the "translation error".

 

Since your code does not explicitly state a reference level it may be that the order of the values is considered differently in English than Chinese since the variable is text and the order of text variables could be considerably different based on sort order of the specific language/ character set.

 

When in doubt specify the exact value you want for your reference levels.

 

 

Tom
Super User
Super User

Looks like it is using the same value for the Reference since the OR is the same.

Is it printing that row heading in English?

Looks like it has change the way it is generating the row description, instead of listing Comparator VS Reference it is listing Reference-Comparator? I have no idea why it would do that. 

Are you sure you are on the same version of SAS in both environments?

What value do you see for the SYSVLONG automatic macro variable?

 

Slash
Quartz | Level 8

The default reference level is the last ordered level, the order of the values in English is same as in Chinese. So the ordered value is "Asia", "Europe", and "USA". I can explicitly specify the reference level, the results remain the same.

 

When reference level is "Asia":

 

data test;
    set sashelp.cars;
    where DriveTrain in ('Front', 'Rear');
run;

proc logistic data=test plots(only)=all;
    class origin(param=ref ref='Asia');
    model DriveTrain (event = 'Front')=msrp enginesize horsepower mpg_highway origin;
run;

 

Here is the partial results in English version:

EnglishVersion2.png

 

Here is the partial results in Chinese version:

ChineseVersion2.png

 

Change the reference level to 'Europe':

 

data test;
    set sashelp.cars;
    where DriveTrain in ('Front', 'Rear');
run;

proc logistic data=test plots(only)=all;
    class origin(param=ref ref='Europe');
    model DriveTrain (event = 'Front')=msrp enginesize horsepower mpg_highway origin;
run;

Here is the partial results in English version:

 

EnglishVersion3.png

 

Here is the partial results in Chinese version:

ChineseVersion3.png

 

So, I think this makes me very confused before I run the code in English version.

 

As Tom said, it has change the way it is generating the row description, instead of listing Comparator VS Reference it is listing Reference-Comparator. And I'm pretty sure I'm  the same version of SAS for different language config.

 

By the way:

SYSVLONG: 9.04.01M5P091317.

OS: Windows 10 Pro (Version:1809, OS Build: 17763.379)

ChrisHemedinger
Community Manager

Hi @Slash - I've passed this information on to our Chinese localization team to investigate.

Slash
Quartz | Level 8

@ChrisHemedinger Thanks. I just think about to mention you at Twitter.Smiley Happy

Tom
Super User
Super User

I suspect that they had a hard time translating the meaning of "VS" and decided it would be better to just not use it, that the hypen would make more sense to Chinese users.

ChrisHemedinger
Community Manager
Status changed to: Under Consideration

Our localization team is looking at the best way to handle this now.  They are tracking it for a fix in a future release.  They recognize that using the "-" (dash) is ambiguous, as it could be interpreted as minus.  So they are considering a Chinese version of "as compared to".

ChrisHemedinger
Community Manager
Status changed to: Suggestion Implemented

I'm marking this as Implemented -- not available quite yet but it is scheduled for a SAS/STAT hotfix very soon.  The localization team decided to simply revert to the English version of this text ("Origin Asia vs USA" in your example) to avoid ambiguity.

 

Thanks for reporting this!

Slash
Quartz | Level 8

@ChrisHemedinger Thanks, good to know.