All, Recenly, I am learning the PROC SEQDESIGN procedure via book Modern Approaches in clinical trials using SAS authored by Sandeep and Zink, but found the columns could not be replicated
The code for creating the last 3 columns are as below:
ods graphics on;
proc seqdesign altref=0.19 errspend pss(cref=0 0.5 1) stopprob(cref=0 0.5 1) plots=(asn power errspend) boundaryscale=stdZ;
OneSidedErrorSpending: design nstages=2 method(alpha)=ERRFUNCGAMMA(GAMMA=-4) ALT=UPPER STOP=REJECT/*REJECT NULL STAND FOR STOP due to efficacy*/
alpha=0.025
beta=0.2
info=cum(1 2);
samplesize model=twosamplemean(stddev=1 weight=1);
run;
ods graphics off;
After running such code, there is no results match the numbers in the shot from the book below.
For instance, the results generated by above code gives the size will be 870 not 799 as shown in the book.
I consulted the authors as well, but seems not reply.
Anybody could help me to figure out this? THank you in advance.
@Jack2012 wrote:
(...) Why set the different beta values. And still can't connect these values. Generally, we specify a fixed power, i.e. 1-beta, then calculate the size and the corresponding "Early stop probability for efficacy at stage 1", but now this logic seems not workable.
To me it looks like the authors tried to replicate the power values (0.80, 0.84, ..., 0.96 -- possibly, rounded to three decimals, these were really 0.799, 0.837, ..., 0.957?) from the "Classical Design with pessimistic d" using the "Group Sequential Design with pessimistic d" in order to obtain "Average Sample Size" values comparable to the (constant) "Total Sample Size" of the classical design.
I think you need to closely at your output.
When I run your code I see this as part of the output. I think you are confusing "Max Sample Size" from your output with "Average Sample size" in the publication. They apparently changed the label from "Expected Sample Size"
Hello @Jack2012,
I'm not an expert on PROC SEQDESIGN, so not sure if this helps:
On page 89 of the book (as seen in Google Books) it says: "Example Code 3.3 on page 89 shows how this can be done in SAS using method (alpha)=ERRFUNCOBF or using method (alpha)=ERRFUNCGAMMA(GAMMA-4) [sic! missing "=" sign before "-4"] in proc seqdesign [32]."
After replacing ERRFUNCGAMMA(GAMMA=-4) in your code with ERRFUNCOBF and using beta=1-Power (from the highlighted part of your screenshot), i.e., beta=0.201 for altref=0.19 (beta=0.163 for altref=0.20, etc.), I get Stopping probability (Stage_1) and Expected Sample Size (Alt Ref) (almost) equal to the values in your table and the same holds for the other altref values (0.20, 0.21, ..., 0.25). The rounding error in beta might explain the remaining small differences.
@Jack2012 wrote:
(...) Why set the different beta values. And still can't connect these values. Generally, we specify a fixed power, i.e. 1-beta, then calculate the size and the corresponding "Early stop probability for efficacy at stage 1", but now this logic seems not workable.
To me it looks like the authors tried to replicate the power values (0.80, 0.84, ..., 0.96 -- possibly, rounded to three decimals, these were really 0.799, 0.837, ..., 0.957?) from the "Classical Design with pessimistic d" using the "Group Sequential Design with pessimistic d" in order to obtain "Average Sample Size" values comparable to the (constant) "Total Sample Size" of the classical design.
Are you ready for the spotlight? We're accepting content ideas for SAS Innovate 2025 to be held May 6-9 in Orlando, FL. The call is open until September 16. Read more here about why you should contribute and what is in it for you!
ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.
Find more tutorials on the SAS Users YouTube channel.