Slice by statement for a continuous variable

[KKIND]

Here is the code I have for categorical independent variables, comparing across all levels of each variable.

proc logistic data = NEDS.neds_2016_pneumonia_recat;
class died_visit (ref = "0") pay1recat (ref = "3") hosp_traumarecat (ref = "0") / param = glm;
model died_visit = pay1recat hosp_traumarecat pay1recat*hosp_traumarecat / link = glogit;
slice pay1recat*hosp_traumarecat/sliceby(pay1recat='3') diff oddsratio cl;
slice pay1recat*hosp_traumarecat/sliceby(pay1recat='2') diff oddsratio cl;
slice pay1recat*hosp_traumarecat/sliceby(pay1recat='1') diff oddsratio cl;
slice pay1recat*hosp_traumarecat/sliceby(hosp_traumarecat='0') diff oddsratio cl;
slice pay1recat*hosp_traumarecat/sliceby(hosp_traumarecat='1') diff oddsratio cl;
slice pay1recat*hosp_traumarecat/sliceby(hosp_traumarecat='2') diff oddsratio cl;
slice pay1recat*hosp_traumarecat/sliceby(hosp_traumarecat='3') diff oddsratio cl;
run;

Is the following code okay for a continuous variable such as age?

slice pay1recat*age/sliceby(pay1recat='3') diff oddsratio cl;
slice pay1recat*age/sliceby(pay1recat='2') diff oddsratio cl;
slice pay1recat*age/sliceby(pay1recat='1') diff oddsratio cl;

Thank you!

[sbxkoenk · Accepted solution]

Have just moved this to "Statistical Procedures" - board.

[KKIND]

Thank you sbxkoenk!

[sbxkoenk]

---

@KKIND wrote:

Is the following code okay for a continuous variable such as age?

slice pay1recat*age/sliceby(pay1recat='3') diff oddsratio cl;
slice pay1recat*age/sliceby(pay1recat='2') diff oddsratio cl;
slice pay1recat*age/sliceby(pay1recat='1') diff oddsratio cl;

Thank you!

---

No.
You will get an ERROR.
ERROR: Only CLASS variables allowed in this effect.

See
SAS/STAT® 15.2 User's Guide
Shared Concepts and Topics
SLICE Statement
https://go.documentation.sas.com/doc/en/statug/15.2/statug_introcom_sect064.htm

It says :
"The SLICE statement is similar to the LSMEANS statement. You use it to perform inferences on model effects that consist entirely of classification variables. With the SLICE statement, these effects must be higher-order effects of at least two classification variables."

Koen

[KKIND]

Thanks again Koen!

[StatDave]

You don't need the SLICE statement to assess the effect of AGE at each level of the interacting categorical predictor. You can do that using the ODDSRATIO statement. For example:

oddsratio age;

BTW, this note discusses assessing the effect of a continuous variable in any generalized linear model. As noted there, particularly in the example of a continuous-categorical interaction in a log-linked model, you can also use the Margins macro. That macro is used to both fit the model and provided the assessment.

[KKIND]

Thanks much StatDave!