BGLIMM

ksaikali — Fri, 20 Oct 2023 14:54:21 GMT

proc bglimm data=ADQS nmc=100000 thin=2 seed=976352
outpost=CenterOut;
class GPPpGaBL TRTP;
model resp(event='1') = gpppgabl trtp/ dist=bin link=logit;
*random int / subject=Center monitor;
estimate "log OR" TRTP 0 1 -1/e;
ods output estimates = est;
run;

data est;
set est ;
odds750=exp(Mean);
o750l=exp(HPDLower);
o750u=exp(HPDUpper);
run;

data centerout;
set centerout;

p_750 = logistIc(intercept +GPPPGABL__3__MODERATE+TRTP2);
p_placebo=logistic(intercept +GPPPGABL__3__MODERATE);
d_750=p_750-p_placebo;
r_750=p_750/p_placebo;
o_750=(p_750/(1-p_750))/(p_placebo/(1-p_placebo));

if d_750 < 0 then pd_750=2 ; else pd_750=0;

if o_750<1 then po_750=2 ; else po_750=0;

run;

1st question , can I compute the odds ratio using o_750 calculations ?

2ns question ; why I am getting different results between the mean of o_750 and odds750 from the estimate statement?

topic BGLIMM in Statistical Procedures

BGLIMM