topic Re: How to output oddsratio and parameterestimates from interaction terms in proc logistic in Statistical Procedures

How to output oddsratio and parameterestimates from interaction terms in proc logistic

TineKopp — Mon, 09 Dec 2019 09:50:09 GMT

I have imputed missing values in my dataset by proc mi and wants to provide odds ratios and parameterestimates in seperate datasets on the interaction between a categorical (age, two levels) and continuous (Att24_pa: number of relapses per year) variable. However, when I run proc logistic with an ODDSRATIO statement followed by ods out, I only get datasets with odds ratios and parameterestimates on the variables not included in the interactionterm (here: sex, diseaseDurationYrs and baselineedss). "_mi_fcs_komplet" is the dataset with imputed missing values. So, when I run this model, I get one odds ratio from each level in the age category for the increment in the continuous Att24_pa variable. And I need to output these two estimates with confidence intervals so I can combine them using proc mianalyze, followed by log-transforming, combining again and, finally, back-transformation. Also, I'm not sure whether the proc mianayze statement is correct for interaction terms.

This is the code I have used:

proc logistic data=_mi_fcs_komplet ;
by _Imputation_; *100 imputations;
class Efficacy (ref='No') age (ref=">=40") sex (ref="2 Female") ;
model Efficacy = age|Att24_pa sex diseaseDurationYrs baselineedss ;
oddsratio att24_pa ; *This is the information I need to be outputted;
ODS OUTPUT PARAMETERESTIMATES=lgsparms ODDSRATIOS=lgsodds;
run;

/* Combining the 100 imputations */
proc mianalyze parms(classvar=CLASSVAL )=lgsparms;
class age sex;
modeleffects age|Att24_pa sex diseaseDurationYrs baselineedss;
ODS OUTPUT PARAMETERESTIMATES=mian_ /*lgsparms*/;
run;
*** Log-transform odds ratio estimates and obtain standard error from confidence intervals ***;
*Får dog kun alderseffekten ud her;
DATA lgsodds_t; SET lgsodds(WHERE=(INDEX(EFFECT,"att24_pa")));
log_or_lr_value=LOG(ODDSRATIOEST);
log_or_lr_se=(LOG(UPPERCL)-LOG(LOWERCL))/(2*1.96);
RUN;
*** Combine transformed estimates;
PROC MIANALYZE DATA=lgsodds_t;
ODS OUTPUT PARAMETERESTIMATES=mian_lgsodds_t;
MODELEFFECTS log_or_lr_value;
STDERR log_or_lr_se;
RUN;
*** Back-transform combined values;
DATA mian_lgsodds_bt; SET mian_lgsodds_t;
Estimate_back = EXP(ESTIMATE); *Pooled odds ratio;
LCL_back=Estimate_back*EXP(-1.96*STDERR); *Pooled lower limit;
UCL_back=Estimate_back*EXP(+1.96*STDERR); *Pooled upper limit;
RUN;

Any help would be much appreciated.

Many Thanks,

Tine

Re: How to output oddsratio and parameterestimates from interaction terms in proc logistic

Reeza — Mon, 09 Dec 2019 17:35:36 GMT

Here's some instructions and explanations on how to capture output that is shown.
https://blogs.sas.com/content/sastraining/2017/03/31/capturing-output-from-any-procedure-with-an-ods-output-statement/

@TineKopp wrote:

Hi

I have imputed missing values in my dataset by proc mi and wants to provide odds ratios and parameterestimates in seperate datasets on the interaction between a categorical (age, two levels) and continuous (Att24_pa: number of relapses per year) variable. However, when I run proc logistic with an ODDSRATIO statement followed by ods out, I only get datasets with odds ratios and parameterestimates on the variables not included in the interactionterm (here: sex, diseaseDurationYrs and baselineedss). "_mi_fcs_komplet" is the dataset with imputed missing values. So, when I run this model, I get one odds ratio from each level in the age category for the increment in the continuous Att24_pa variable. And I need to output these two estimates with confidence intervals so I can combine them using proc mianalyze, followed by log-transforming, combining again and, finally, back-transformation. Also, I'm not sure whether the proc mianayze statement is correct for interaction terms.

This is the code I have used:

proc logistic data=_mi_fcs_komplet ;
by _Imputation_; *100 imputations;
class Efficacy (ref='No') age (ref=">=40") sex (ref="2 Female") ;
model Efficacy = age|Att24_pa sex diseaseDurationYrs baselineedss ;
oddsratio att24_pa ; *This is the information I need to be outputted;
ODS OUTPUT PARAMETERESTIMATES=lgsparms ODDSRATIOS=lgsodds;
run;

/* Combining the 100 imputations */
proc mianalyze parms(classvar=CLASSVAL )=lgsparms;
class age sex;
modeleffects age|Att24_pa sex diseaseDurationYrs baselineedss;
ODS OUTPUT PARAMETERESTIMATES=mian_ /*lgsparms*/;
run;
*** Log-transform odds ratio estimates and obtain standard error from confidence intervals ***;
*Får dog kun alderseffekten ud her;
DATA lgsodds_t; SET lgsodds(WHERE=(INDEX(EFFECT,"att24_pa")));
log_or_lr_value=LOG(ODDSRATIOEST);
log_or_lr_se=(LOG(UPPERCL)-LOG(LOWERCL))/(2*1.96);
RUN;
*** Combine transformed estimates;
PROC MIANALYZE DATA=lgsodds_t;
ODS OUTPUT PARAMETERESTIMATES=mian_lgsodds_t;
MODELEFFECTS log_or_lr_value;
STDERR log_or_lr_se;
RUN;
*** Back-transform combined values;
DATA mian_lgsodds_bt; SET mian_lgsodds_t;
Estimate_back = EXP(ESTIMATE); *Pooled odds ratio;
LCL_back=Estimate_back*EXP(-1.96*STDERR); *Pooled lower limit;
UCL_back=Estimate_back*EXP(+1.96*STDERR); *Pooled upper limit;
RUN;

Any help would be much appreciated.

Many Thanks,

Tine

Re: How to output oddsratio and parameterestimates from interaction terms in proc logistic

TineKopp — Wed, 11 Dec 2019 08:23:47 GMT

Dear Reeza

Thank you for the link. However, I have already used the ods output statement so the question is what I should call the output object name in order to output odds ratios from the oddsratio statement. I have managed to output the interaction term, but I still need the odds ratios produced by my statement "oddsratio att24_pa" which is the odds ratio of Att24_pa at age=<40 and Att24_pa at age=>=40.

Many thanks,

Tine

Re: How to output oddsratio and parameterestimates from interaction terms in proc logistic

TineKopp — Wed, 11 Dec 2019 18:15:55 GMT

Dear Reeza

I just found out that the statement ods output OddsRatiosWald=outputdata provides me with oddsratios from the oddsratio statement. The final model should look like this:

proc logistic data=_mi_fcs_komplet ;
by _Imputation_; *100 imputations;
class Efficacy (ref='No') age (ref=">=40") sex (ref="2 Female") ;
model Efficacy = age|Att24_pa sex diseaseDurationYrs baselineedss ;
oddsratio att24_pa ;
ODS OUTPUT PARAMETERESTIMATES=lgsparms ODDSRATIOS=lgsodds;

ODS OUTPUT OddsRatiosWald=orw;*These are the odds ratios from the oddsratio statement;

run;

Another solution could be to use dummyvariables for the interaction terms. That work just as well!

Many thanks,

Tine

Re: How to output oddsratio and parameterestimates from interaction terms in proc logistic

Reeza — Wed, 11 Dec 2019 18:17:32 GMT

Did you follow the instructions in the blog post - put ODS TRACE ON before/after your code and after and see what the table name is from the log or output? It's the first step.

You can have multiple ODS OUTPUT statements.

https://statgeeks.wordpress.com/2013/08/

Or check the documentation for table names.

https://documentation.sas.com/?docsetId=statug&docsetVersion=15.1&docsetTarget=statug_logistic_details112.htm&locale=en

Most likely it will be Oddsratios, oddsratiosPL or OddsratiosWALD from the table.

@TineKopp wrote:

Dear Reeza

Thank you for the link. However, I have already used the ods output statement so the question is what I should call the output object name in order to output odds ratios from the oddsratio statement. I have managed to output the interaction term, but I still need the odds ratios produced by my statement "oddsratio att24_pa" which is the odds ratio of Att24_pa at age=<40 and Att24_pa at age=>=40.

Many thanks,

Tine

Re: How to output oddsratio and parameterestimates from interaction terms in proc logistic

TineKopp — Wed, 11 Dec 2019 18:23:12 GMT

Yes, thank you! And as I just wrote. If you dont know which table name to use, make dumme variables instead. That gives the same results.

Best whishes,

Tine

Re: How to output oddsratio and parameterestimates from interaction terms in proc logistic

Reeza — Wed, 11 Dec 2019 18:56:36 GMT

Only in this particular example, in general for output the approach above works on any type of output from SAS PROCS.