Invalid reference value for nationality

[zalahmed]

how to solve this error below? nationality variables value either 1 or 2 in the data set , 

SAS code: 

proc logistic data=work.project1 plots=ODDSRATIO

 outest=betas covout ; 

 class sector (param=ref ref='4') age_cat (param=ref ref='1') speciality_cat(param=ref ref='1') gender(param=ref ref='1')

nationality (param=ref ref='2') education_cat (param=ref ref='1')  university_cat(param=ref ref='1') number_adult_cat(param=ref ref='2')

number_nurses_cat(param=ref ref='1')BMI_cat(param=ref ref='1') physician_hgeneral_cat(param=ref ref='1') physician_chronic(param=ref ref='2') 

smoking(param=ref ref='1') sleep_hours_cat(param=ref ref='1') knowledge_all_cat(param=ref ref='1')

aerobic_cat(param=ref ref='1') marital_cat(param=ref ref='1') YWEG_cat(param=ref ref='1')

number_pediatric_cat(param=ref ref='2') number_physicians_cat(param=ref ref='1')taining (param=ref ref='2') referal_system(param=ref ref='2');       

model provide_general_without_cat(Event='2')=sector age_cat speciality_cat gender nationality education_cat university_cat number_adult_cat 

number_nurses_cat BMI_cat physician_hgeneral_cat physician_chronic smoking sleep_hours_cat 

knowledge_all_cat  marital_cat YWEG_cat number_pediatric_cat number_physicians_cat taining 

referal_system  marital_cat YWEG_cat aerobic_cat

/link=logit

selection=stepwise

                    slentry=0.2

                    slstay=0.25

                    details

                    lackfit;

    output out=pred p=phat lower=lcl upper=ucl ;

run;

SAS log:

596

597  proc logistic data=work.project1 plots=ODDSRATIO

598   outest=betas covout ;

599   class sector (param=ref ref='2') age_cat (param=ref ref='1') speciality_cat(param=ref

599! ref='1') gender(param=ref ref='1')

600  nationality (param=ref ref='2') education_cat (param=ref ref='1')  university_cat(param=ref

600! ref='1') number_adult_cat(param=ref ref='2')

601  number_nurses_cat(param=ref ref='1')BMI_cat(param=ref ref='1')

601! physician_hgeneral_cat(param=ref ref='1') physician_chronic(param=ref ref='2')

602  smoking(param=ref ref='1') sleep_hours_cat(param=ref ref='1') knowledge_all_cat(param=ref

602! ref='1')

603  aerobic_cat(param=ref ref='1') marital_cat(param=ref ref='1') YWEG_cat(param=ref ref='1')

604  number_pediatric_cat(param=ref ref='2') number_physicians_cat(param=ref ref='1')taining

604! (param=ref ref='2') referal_system(param=ref ref='2');

605  model provide_general_without_cat(Event='2')=sector age_cat speciality_cat gender nationality

605! education_cat university_cat number_adult_cat

606  number_nurses_cat BMI_cat physician_hgeneral_cat physician_chronic smoking sleep_hours_cat

607  knowledge_all_cat  marital_cat YWEG_cat number_pediatric_cat number_physicians_cat taining

608  referal_system  marital_cat YWEG_cat aerobic_cat

609                      /link=logit

610                      selection=stepwise

611                      slentry=0.2

612                      slstay=0.25

613                      details

614                      lackfit;

615      output out=pred p=phat lower=lcl upper=ucl ;

616  run;

ERROR: Invalid reference value for nationality.

NOTE: The SAS System stopped processing this step because of errors.

NOTE: There were 147 observations read from the data set WORK.PROJECT1.

WARNING: The data set WORK.BETAS may be incomplete.  When this step was stopped there were 0

         observations and 0 variables.

WARNING: Data set WORK.BETAS was not replaced because this step was stopped.

WARNING: The data set WORK.PRED may be incomplete.  When this step was stopped there were 0

         observations and 0 variables.

WARNING: Data set WORK.PRED was not replaced because this step was stopped.

NOTE: PROCEDURE LOGISTIC used (Total process time):

      real time           0.01 seconds

      cpu time            0.01 seconds

[Reeza]

Does nationality have a format applied? The ref option requires the formatted values. 

FYI you can include PARAM=REF after your class statement if all categorical variables will be using it. This means you can reduce your code. 

CLASS ....  / Param=ref;

[zalahmed]

no format applied to nationality 

[Reeza]

Then post a proc freq for the variable as well as a proc contents for your data set. 

[zalahmed]

Attachedproc freq and proc contents  

[Reeza]

I'm not seeing anything specific. What happens if you reduce the model to only nationality? 

You know that you have two many variables for your data? A quick rule of thumb is that you need approximately 30 obs per variable. 

[ballardw]

By any chance to you have a combination of the class variables that does not include nationality='2'? Or is nationality numeric and not character?

[StatDave]

If the variable has a user-defined format assigned, you must specify the formatted value in the REF= option.  Also, note that you do not need to specify PARAM=REF for each variable.  You can specify it once, after a slash, at the end of the CLASS statement. For example: 

class a(ref="1") b(ref="hi") / param=ref;

[Reeza]

@zalahmed at this point you'll either need to provide your data so we can replicate the issue or contact SAS Tech Support for help. I suspect the issue is for a specific level there is no values of 2, but I would expect a different error.

As I previously mentioned, you have too many variables and I don't think it's possible to fit your model. The number of observations must be greater than the number of variables for the matrix math to work out - back to the really basic linear algebra interpretation of regression. 

[zalahmed]

I will try to change the model , and let you know what's work 

Thanks 

[rasmuslarsen]

It looks like this bug:

http://support.sas.com/kb/13/510.html