Not applicable
Posts: 1

# Missing boundary using Proc SEQDESIGN

Hi all,

Has anyone used Proc SEQDESIGN to calculate the boundary values of a sequential group design using O'Brien Fleming type error spending function? I tried the following codes using 11 stages but the boundary values are missing for the first stage. I think this is caused by the information proportion less than 0.1 because when I tried nstage=10 or 12, the procedure produced +/-6.99 and missing values, respectively, for the first stage. The corresponding information proportion is 0.1 and 0.0833, respectively.

codes:

proc seqdesign ;

TwoSidedOBF: design nstages=11 method=errfuncobf ;

run;

The resulting boundary table:

 Boundary Information (Standardized Z Scale) Null Reference = 0 _Stage_ Information Alternative Boundary Values Level Reference Lower Upper Proportion Lower Upper Alpha Alpha 1 0.0909 -0.9952 0.9952 . . 2 0.1818 -1.40742 1.40742 -5.12754 5.12754 3 0.2727 -1.72373 1.72373 -4.13644 4.13644 4 0.3636 -1.9904 1.9904 -3.54589 3.54589 5 0.4545 -2.22533 2.22533 -3.14865 3.14865 6 0.5455 -2.43773 2.43773 -2.85973 2.85973 7 0.6364 -2.63305 2.63305 -2.63784 2.63784 8 0.7273 -2.81485 2.81485 -2.46062 2.46062 9 0.8182 -2.98559 2.98559 -2.31493 2.31493 10 0.9091 -3.14709 3.14709 -2.1924 2.1924 11 1 -3.3007 3.3007 -2.08746 2.08746

The problem only appear when I use 'errfuncobf' (for example, 'obf' and 'errfuncpoc' are both fine and can produce complete boundary values for any stage). Is there any solution to solve this problem and let the procedure produce the value for the first stage?

I am using SAS v9.3.

Thank you.

Discussion stats
• 0 replies
• 171 views
• 0 likes
• 1 in conversation