The SEQDESIGN Procedure

Example 103.8 Creating a One-Sided Error Spending Design

(View the complete code for this example.)

This example requests a five-stage, one-sided group sequential design for normally distributed statistics. The design uses an O’Brien-Fleming-type error spending function for the boundary and a Pocock-type error spending function for the boundary. The following statements request a one-sided design by using different and spending functions:

ods graphics on;
proc seqdesign altref=0.2  errspend
               pss(cref=0 0.5 1)
               stopprob(cref=0 0.5 1)
               plots=(asn power errspend)
               ;
   OneSidedErrorSpending: design nstages=5
                          method(alpha)=errfuncobf
                          method(beta)=errfuncpoc
                          alt=upper  stop=both
                          alpha=0.025
                          ;
run;

The "Design Information" table in Output 103.8.1 displays design specifications and the derived statistics. With the specified alternative reference, the maximum information is derived.

Output 103.8.1: Error Spending Method Design Information

The SEQDESIGN Procedure
Design: OneSidedErrorSpending

Design Information
Statistic DistributionNormal
Boundary ScaleStandardized Z
Alternative HypothesisUpper
Early StopAccept/Reject Null
MethodError Spending
Boundary KeyBoth
Alternative Reference0.2
Number of Stages5
Alpha0.025
Beta0.1
Power0.9
Max Information (Percent of Fixed Sample)119.4278
Max Information313.7196
Null Ref ASN (Percent of Fixed Sample)50.35408
Alt Ref ASN (Percent of Fixed Sample)78.77223


The "Method Information" table in Output 103.8.2 displays the and errors, alternative reference, and derived drift parameter, which is the standardized alternative reference at the final stage.

Output 103.8.2: Method Information

Method Information
BoundaryMethodAlphaBetaError SpendingAlternative
Reference
Drift
Function
Upper AlphaError Spending0.02500.Approx O'Brien-Fleming0.23.542426
Upper BetaError Spending.0.10000Approx Pocock0.23.542426


With the STOPPROB option, the "Expected Cumulative Stopping Probabilities" table in Output 103.8.3 displays the expected stopping stage and cumulative stopping probability to reject the null hypothesis at each stage under various hypothetical references , where is the alternative reference and are values specified in the CREF= option.

Output 103.8.3: Stopping Probabilities

Expected Cumulative Stopping Probabilities
Reference = CRef * (Alt Reference)
CRefExpected
Stopping Stage
SourceStopping Probabilities
Stage_1Stage_2Stage_3Stage_4Stage_5
0.00002.108Reject Null0.000000.000390.003810.012210.02500
0.00002.108Accept Null0.380800.691330.861620.941700.97500
0.00002.108Total0.380800.691730.865430.953911.00000
0.50003.296Reject Null0.000020.012650.096500.244650.38724
0.50003.296Accept Null0.136650.280630.410800.522300.61276
0.50003.296Total0.136670.293280.507300.766951.00000
1.00003.298Reject Null0.000500.132090.526420.803900.90000
1.00003.298Accept Null0.029540.052310.070850.086480.10000
1.00003.298Total0.030040.184400.597280.890391.00000


With the PSS option, the "Power and Expected Sample Sizes" table in Output 103.8.4 displays powers and expected sample sizes under various hypothetical references , where is the alternative reference and are the default values in the CREF= option.

Output 103.8.4: Power and Expected Sample Size Information

Powers and Expected Sample Sizes
Reference = CRef * (Alt Reference)
CRefPowerSample Size
Percent
Fixed-Sample
0.00000.0250050.3541
0.50000.3872478.7219
1.00000.9000078.7722


With the PLOTS=ASN option, the procedure displays a plot of expected sample sizes under various hypothetical references, as shown in Output 103.8.5. By default, expected sample sizes under the hypotheses , , are displayed, where is the alternative reference.

Output 103.8.5: ASN Plot

ASN Plot


With the PLOTS=POWER option, the procedure displays a plot of the power curves under various hypothetical references for all designs simultaneously, as shown in Output 103.8.6. By default, the option CREF= and powers under hypothetical references are displayed, where are values specified in the CREF= option. These CREF= values are displayed on the horizontal axis.

Under the null hypothesis, , the power is 0.025, the upper Type I error probability. Under the alternative hypothesis, , the power is 0.9, one minus the Type II error probability. The plot shows only minor difference between the two designs.

Output 103.8.6: Power Plot

Power Plot


The "Boundary Information" table in Output 103.8.7 displays information level, alternative reference, and boundary values. By default (or equivalently if you specify BOUNDARYSCALE=STDZ), the alternative reference and boundary values are displayed with the standardized Z scale. That is, the resulting standardized alternative reference at stage k is given by , where is the specified alternative reference and is the information level at stage k, .

Output 103.8.7: Boundary Information

Boundary Information (Standardized Z Scale)
Null Reference = 0
_Stage_ AlternativeBoundary Values
Information LevelReferenceUpper
ProportionActualUpperBetaAlpha
10.200062.743931.58422-0.303384.87688
20.4000125.48792.240430.416673.35706
30.6000188.23182.743950.971652.67766
40.8000250.97573.168441.436272.26535
51.0000313.71963.542431.875221.87522


With ODS Graphics enabled, a detailed boundary plot with the rejection and acceptance regions is displayed, as shown in Output 103.8.8. This plot displays the boundary values in the "Boundary Information" table in Output 103.8.7.

Output 103.8.8: Boundary Plot

Boundary Plot


The "Error Spending Information" table in Output 103.8.9 displays cumulative error spending at each stage for each boundary.

Output 103.8.9: Error Spending Information

Error Spending Information
_Stage_Information
Level
Cumulative Error Spending
Upper
ProportionBetaAlpha
10.20000.029540.00000
20.40000.052310.00039
30.60000.070850.00381
40.80000.086480.01221
51.00000.100000.02500


With the PLOTS=ERRSPEND option, the procedure displays a plot of error spending for each boundary, as shown in Output 103.8.10. This plot displays the cumulative error spending at each stage in the "Error Spending Information" table in Output 103.8.9. The O’Brien-Fleming-type spending function is conservative in early stages because it uses much less at early stages than in the later stages. In contrast, the Pocock-type spending function uses more at early stages than in the later stages.

Output 103.8.10: Error Spending Plot

Error Spending Plot