topic Re: negative binomial regression in Statistical Procedures

negative binomial regression

KKIND — Fri, 03 May 2024 22:05:26 GMT

What is the equivalent of slice statement in a negative binomial regression model?

Re: negative binomial regression

Ksharp — Sat, 04 May 2024 01:28:53 GMT

Did you try SLICE statement of PROC GENMOD ?

another way is using LSMEANS + AT option

and another way is using EFFECTPLOT statement like:

data insure;
input n c car$ age;
carnum=1;
if car="medium" then carnum=2;
if car="large" then carnum=3;
ln = log(n);
datalines;
500 42 small 1
1200 37 medium 1
100 1 large 1
400 101 small 2
500 73 medium 2
300 14 large 2
;
proc genmod data=insure;
class  age;
model c = carnum age / dist=NEGBIN link=log ;
effectplot slicefit(x=carnum sliceby=age) / clm ;
run;

Re: negative binomial regression

KKIND — Sun, 05 May 2024 18:44:16 GMT

@Ksharp, many thanks! I will try your sample code and revert back asap. Have a great day!

Best

KKIND

Re: negative binomial regression

KKIND — Sun, 05 May 2024 19:14:37 GMT

when I used the slice statement with multinomial logistic regression as follows.

C:\Users\kiran\OneDrive\HCUP-DBs\NEDS_2019\";

proc logistic data = NEDS.neds_2019_pneumonia_recat;

class died_visit (ref = "1") pay1recat (ref = "3") hosp_controlrecat (ref = "4") pl_nchsrecat (ref = "1") / param = glm;

model died_visit = pay1recat hosp_controlrecat pay1recat*hosp_controlrecat aweekend female pl_nchsrecat / link = glogit;

slice pay1recat*hosp_controlrecat/sliceby(pay1recat='3') diff oddsratio cl;

slice pay1recat*hosp_controlrecat/sliceby(pay1recat='2') diff oddsratio cl;

slice pay1recat*hosp_controlrecat/sliceby(pay1recat='1') diff oddsratio cl;

slice pay1recat*hosp_controlrecat/sliceby(hosp_controlrecat='1') diff oddsratio cl;

slice pay1recat*hosp_controlrecat/sliceby(hosp_controlrecat='2') diff oddsratio cl;

slice pay1recat*hosp_controlrecat/sliceby(hosp_controlrecat='3') diff oddsratio cl;

slice pay1recat*hosp_controlrecat/sliceby(hosp_controlrecat='4') diff oddsratio cl;

run;

one of the output tables I got and was looking for is

Per my research question, I am looking at within hospital variation in outcomes by payer type (1=Medicare; 2=Medicaid; and 3= private insurance). From this output above, I can determine what is the difference in outcomes comparing between each payer type for each level of the hospital ownership type (hosp_controlrecat). Is there some way I can get similar outputs (comparing between payer types) under the proc genmod code? Thank you!

Best

KKIND

Re: negative binomial regression

Ksharp — Mon, 06 May 2024 02:02:29 GMT

Did you try SLICE statement?

data insure;
input n c car$ age;
carnum=1;
if car="medium" then carnum=2;
if car="large" then carnum=3;
ln = log(n);
datalines;
500 42 small 1
1200 37 medium 1
100 1 large 1
400 101 small 2
500 73 medium 2
300 14 large 2
;
ods select  SliceDiffs;
proc genmod data=insure;
class  carnum age;
model c = carnum|age / dist=NEGBIN link=log offset=ln;
slice carnum*age /sliceby=age diff cl  ;
run;

Or @StatDave could give you a hand.

Re: negative binomial regression

KKIND — Tue, 07 May 2024 15:03:29 GMT

Dear Ksharp,

Thanks again for your efforts and sharing the code. I tried using it bit still couldn't get the output as you shared.

Best

KKind

Re: negative binomial regression

KKIND — Fri, 07 Jun 2024 15:27:18 GMT

Dear Ksharp and all,

Thank you for your advice on the slice statement for the negative binomial regression statement . I have managed to find the correct code to get the comparison I was looking for. Here is the code and one of the output tables.

proc genmod data=NIS.nis_2019_septicemia_ipexcmicat;

class pay1recat (ref = "3") hosp_region (ref= "4") hosp_locteach (ref= "1") hosp_bedsize (ref= "1") h_contrlrecat (ref= "3") female (ref= "0") zipinc_qrtl (ref= "4") aprdrg_severity (ref= "4") racerecat (ref= "1") cmr_index_mortalitycat (ref= "1") / param=glm;

model los = pay1recat hosp_region pay1recat*hosp_region hosp_locteach pay1recat*hosp_locteach hosp_bedsize pay1recat*hosp_bedsize h_contrlrecat pay1recat*h_contrlrecat female zipinc_qrtl aprdrg_severity racerecat cmr_index_mortalitycat / dist=negbin link=log;

lsmeans pay1recat*hosp_region / adjust=tukey;

slice pay1recat*hosp_region / sliceby=hosp_region diff cl;

lsmeans pay1recat*hosp_locteach / adjust=tukey;

slice pay1recat*hosp_locteach / sliceby=hosp_locteach diff cl;

lsmeans pay1recat*hosp_bedsize / adjust=tukey;

slice pay1recat*hosp_bedsize / sliceby=hosp_bedsize diff cl;

lsmeans pay1recat*h_contrlrecat / adjust=tukey;

slice pay1recat*h_contrlrecat / sliceby=h_contrlrecat diff cl;

run;

Re: negative binomial regression

Ksharp — Sat, 08 Jun 2024 06:26:41 GMT

It is better to post your question at STAT forum:

https://communities.sas.com/t5/Statistical-Procedures/bd-p/statistical_procedures

and calling @StatDave @SteveDenham

Re: negative binomial regression

KKIND — Sun, 09 Jun 2024 00:56:05 GMT

Hello Ksharp,

It is not a question > I was just updating the status of my question on teh slice statement for negative binomial regression that I had posed, informing that I have found teh solution and I pasted the solution and the corresponding output. Thanks!

Re: negative binomial regression

SteveDenham — Thu, 13 Jun 2024 13:54:10 GMT

Given that, you should mark @Ksharp 's post where he suggests a SLICE statement as a correct answer.

SteveDenham