PROC SEQTEST Statement |
Table 81.1 summarizes the options in the PROC SEQTEST statement.
Option |
Description |
---|---|
Input Data Sets |
|
Specifies the data set for boundary information |
|
Specifies the data set for parameter estimates and information levels |
|
Specifies the data set for parameter estimates and standard errors |
|
Boundaries |
|
Checks for overlapping of the lower and upper boundaries |
|
at the current and subsequent interim stages in a two-sided design |
|
Specifies the boundary key to maintain Type I and II error probability levels |
|
Specifies the boundary scale |
|
Specifies error spending methods for boundary adjustments |
|
Specifies minimum error spending values for the boundaries |
|
Specifies whether information levels at future interim stages should be |
|
adjusted |
|
Specifies the number of stages |
|
|
|
Specifies the test variable in the DATA= data set |
|
Specifies the test variable in the PARMS= data set |
|
-Values and Confidence Intervals |
|
Specifies the significance levels for the confidence interval |
|
Specifies the types of confidence interval |
|
Specifies the ordering of the sample space used to derive |
|
the -values and confidence limits |
|
Table Output |
|
Displays conditional powers |
|
Displays the cumulative error spending at each stage |
|
Displays the predictive powers |
|
Displays the powers and expected sample sizes |
|
Displays the repeated confidence intervals |
|
Displays the expected cumulative stopping probabilities |
|
Graphics Output |
|
Displays the expected sample numbers plot |
|
Displays the conditional powers plot |
|
Displays the error spending plot |
|
Displays the powers plot |
|
Displays the repeated confidence intervals plot |
|
Displays the boundary plot with test statistics |
The BOUNDARY= option provides the information for the design and is required in the PROC SEQTEST statement. By default, the SEQTEST procedure displays tables of design information and test information. If ODS Graphics is enabled, the procedure also displays a sequential test plot.
The following options can be used in the PROC SEQTEST statement. They are listed in alphabetical order.
specifies whether to check for overlapping of the lower and upper boundaries for the two corresponding one-sided tests at the current and subsequent interim stages. This option applies to two-sided designs with early stopping to accept , or to either accept or reject . This type of overlapping might result from a small spending at an interim stage. When you specify BETAOVERLAP=ADJUST, the procedure checks for this type of overlapping at the current and subsequent interim stages. If such overlapping is found, the boundaries for the two-sided design at that stage are set to missing, and the spending values at subsequent stages are adjusted, as described in the section Boundary Adjustments for Overlapping Lower and Upper beta Boundaries.
You can specify BETAOVERLAP=NOADJUST to request that no adjustment be made. The default is BETAOVERLAP=ADJUST.
names the required SAS data set that contains the design boundary information. At stage , the data set is usually created from the "Boundary Information" table created by the SEQDESIGN procedure. At each subsequent stage, the data set is usually created from the "Test Information" table created by the SEQTEST procedure at the previous stage. The data set includes the variables _Scale_ for the boundary scale, _Stop_ for the stopping criterion, and _ALT_ for the type of alternative hypothesis. It also includes _Stage_ for the stage number, Info_Prop for the information proportion, and a set of the boundary variables from Bound_LA, Bound_LB, Bound_UB, and Bound_UA for boundary values at each stage.
The data set might also include _Info_ for the actual information level, NObs for the number of observation, and Events for the number of events required at each stage.
specifies the boundary key to be maintained in the boundary adjustments. The BOUNDARYKEY=ALPHA option maintains the Type I level and derives the Type II error probability, and the BOUNDARYKEY=BETA option maintains the Type II level and derives the Type I error probability. The BOUNDARYKEY=BOTH option maintains both and levels simultaneously by deriving a new maximum information. The default is BOUNDARYKEY=ALPHA.
specifies the boundary scale to be displayed in the output boundary table and plot. The BOUNDARYSCALE=MLE, BOUNDARYSCALE=SCORE, BOUNDARYSCALE=STDZ, and BOUNDARYSCALE=PVALUE options correspond to the boundary with the maximum likelihood estimator scale, score statistic scale, standardized normal scale, and -value scale, respectively. The default is BOUNDARYSCALE=STDZ.
With the BOUNDARYSCALE=MLE or BOUNDARYSCALE=SCORE option, either the MAXINFO= option must be specified or the _Info_ variable must be in the BOUNDARY= data set to provide the necessary information level at each stage to derive the boundary values. Usually, these values are obtained from analysis output in SAS procedures.
Note that for a two-sided design, the -value scale displays the one-sided fixed-sample -value under the null hypothesis with a lower alternative hypothesis.
specifies the significance levels for the confidence interval, where , , and . The default is CIALPHA= 0.05.
For a lower confidence interval (CITYPE=LOWER), the CIALPHA= option produces a lower confidence interval. For an upper confidence interval (CITYPE=UPPER), the CIALPHA= option produces a upper confidence interval. The LOWER= and UPPER= suboptions are applicable only for a two-sided confidence interval (CITYPE=TWOSIDED). The LOWER= suboption specifies the lower significance level and the upper significance level . The UPPER= suboption specifies the upper significance level and the lower significance level . If both LOWER= and UPPER= suboptions are not specified, . The significance levels and are then used for the lower confidence limit and upper confidence limit, respectively.
specifies the type of confidence interval. The CITYPE=LOWER, CITYPE=UPPER, and CITYPE=TWOSIDED options correspond to the lower confidence interval, upper confidence interval, and two-sided confidence interval, respectively. The default is CITYPE=LOWER for the design with an upper alternative, CITYPE=UPPER for the design with a lower alternative, and CITYPE=TWOSIDED for the design with a two-sided alternative.
names the SAS data set that contains the test statistic and its associated information level for the stage. The data set includes the stage variable _Stage_ and a variable to identify or derive the information level: _Info_ for the information level, NObs for the number of observation, or Events for the number of events. If the information level that corresponds to the test statistic is not available, the information level derived in the BOUNDARY= data set is used.
If the TESTVAR= option is specified, the data set also includes the test variable specified in the TESTVAR= option and the scale variable _Scale_ for the test statistic. Usually, these test variable values are obtained from analysis output in SAS procedures.
specifies methods to compute the error spending values at the current and future interim stages for the boundaries. This option is applicable only if the observed information level at the current stage does not match the value provided in the BOUNDARY= data set. These error spending values are then used to derive the updated boundary values. The default is ERRSPENDADJ=ERRLINE. Note that the information levels at future interim stages are determined by the INFOADJ= option.
The following options specify available error spending methods for boundary adjustment:
specifies that the cumulative error spending at each interim stage not be changed, even if the corresponding information level has been changed.
specifies the linear interpolation method for the adjustment.
specifies the gamma function method for the adjustment. The GAMMA= suboption specifies the parameter in the function, where . The default is GAMMA=.
specifies the approximate O’Brien-Fleming cumulative error spending function for the adjustment.
specifies the approximate Pocock cumulative error spending function for the adjustment.
specifies the power function method for the adjustment. The RHO= suboption specifies the power parameter in the function, where . The default is RHO=.
See the section Boundary Adjustments for Information Levels for a detailed description of the available error spending methods for boundary adjustment in the SEQTEST procedure.
If an error spending method for boundary adjustments is used for all boundaries in a group sequential test, you can use the ERRSPENDADJ=method option to specify the method. Otherwise, you can use the following ERRSPENDADJ(boundary)=method options to specify different methods for the boundaries.
specifies the adjustment method for the (rejection) boundary of a one-sided design or the lower and upper boundaries of a two-sided design.
specifies the adjustment method for the lower boundary of a two-sided design.
specifies the adjustment method for the upper boundary of a two-sided design.
specifies the adjustment method for the (acceptance) boundary of a one-sided design or the lower and upper boundaries of a two-sided design.
specifies the adjustment method for the lower boundary of a two-sided design.
specifies the adjustment method for the upper boundary of a two-sided design.
specifies the minimum error spending values at the current observed and future interim stages for the boundaries specified in the BOUNDARYKEY= option. The default is ERRSPENDMIN=0.
If a set of numbers is used for each boundary in the design, you can use the ERRSPENDMIN=numbers option. Otherwise, you can use the following ERRSPENDMIN(boundary)=numbers options to specify different sets of minimum error spending values for the boundaries. For a boundary, the error spending value at stage is identical to its nominal -value.
specifies the minimum error spending values for the boundary of a one-sided design or the lower and upper boundaries of a two-sided design.
specifies the minimum error spending values for the lower boundary of a two-sided design.
specifies the minimum error spending values for the upper boundary of a two-sided design.
specifies the minimum error spending values for the boundary of a one-sided design or the lower and upper boundaries of a two-sided design.
specifies the minimum error spending values for the lower boundary of a two-sided design.
specifies the minimum error spending values for the upper boundary of a two-sided design.
specifies whether information levels at future interim stages are to be adjusted. If you specify INFOADJ=NONE, no adjustment is made, and the information levels are preserved at the levels provided in the BOUNDARY= data set. If you specify INFOADJ=PROP (which is the default), the information levels are adjusted proportionally from the levels provided in the BOUNDARY= data set. The section Information Level Adjustments at Future Stages describes how the adjustments are computed.
Note that if you specify BOUNDARYKEY=BOTH, the INFOADJ=NONE option is not applicable, and the INFOADJ=PROP option is used to adjusted the information levels at future stages proportionally from the levels provided in the BOUNDARY= data set to maintain both and levels.
specifies the number of stages for the clinical trial. The default is the number derived from the BOUNDARY= data set.
The specified NSTAGES= number might or might not be the same as the number derived in the BOUNDARY= data set. You can use the NSTAGES= option to set the next stage as the final stage to compute the conditional power, as described in the section Conditional Power Approach.
specifies the ordering of the sample space , where is the stage number and is the observed standardized statistic. The ordering is used to derive the -values for the observed statistic and to create unbiased median estimate and confidence limits from the statistic. The ORDER=LR option specifies the LR ordering that compares the distances between observed standardized statistics and their corresponding hypothetical values, the ORDER=MLE option specifies the MLE ordering that compares values in the MLE scale, and the ORDER=STAGEWISE specifies the stagewise ordering that uses counterclockwise ordering around the continuation region. The default is ORDER=STAGEWISE. See the section Available Sample Space Orderings in a Sequential Test for a detailed description of these sample space orderings.
names the SAS data set that contains the parameter estimate and its associated standard error for the stage. The data set includes the stage variable _Stage_, the test statistic Estimate, the standard error of the estimate StdErr, and the test statistic scale variable _Scale_. The standard error is are used to derive the information level. If the standard error is not available, the information level derived in the BOUNDARY= data set is used.
The data set also includes the variable Parameter, Effect, Variable, or Parm that contains the test variable specified in the TESTVAR= option. Usually, these test variable values are obtained from analysis output in SAS procedures.
The following options can be used in the PROC SEQTEST statement to display additional table output. They are listed in alphabetical order.
displays conditional powers given the most recently observed statistic under specified hypothetical references, where the numbers . In the SEQTEST procedure, the conditional power is the probability that the test statistic at the final stage would exceed the rejection critical value given the observed statistic.
If interim stages exist between the current stage and the final stage, the conditional power is not the conditional probability to reject the null hypothesis . In this case, you can set the next stage as the final stage, and the conditional power is the conditional probability to reject .
For a one-sided test, the powers are derived under the hypothetical references and , where is the observed statistic, is the alternative reference, and are the values specified in the CREF= option. For a two-sided test, the powers are derived under hypothetical references , , and , where is the lower alternative reference and is the upper alternative reference. The default is CREF= 0 0.5 1.0 1.5.
displays the error spending at each stage for each sequential boundary.
displays predictive powers given the most recently observed statistic. The predictive power is the posterior probability that the test statistic at the final stage would exceed the rejection critical value given the observed statistic and a prior distribution of the hypothetical reference. A noninformative prior is used in the procedure.
displays powers and expected sample sizes under various hypothetical references, where the numbers .
For a one-sided design with the null reference , the power and expected sample sizes under hypotheses are displayed, where is the alternative reference and are the values specified in the CREF= option.
For a two-sided design, the power and expected sample sizes under hypotheses and are displayed, where and are the lower and upper alternative references, respectively. The default is CREF= 0 0.5 1.0 1.5.
Note that for a symmetric two-sided design, only the power and expected sample sizes under hypotheses are derived.
displays repeated confidence intervals for the parameter from the observed statistic at each stage. Repeated confidence intervals include both rejection and acceptance confidence intervals.
With the STOP=REJECT or STOP=BOTH option, rejection confidence limits can be derived, and the null hypothesis is rejected if the lower rejection confidence limit is greater than or the upper rejection confidence limit is less than .
With the STOP=ACCEPT or STOP=BOTH option, acceptance confidence limits can be derived, and the null hypothesis is accepted with alternative hypotheses and if the upper acceptance confidence limit is less than and the lower acceptance confidence limit is greater than .
displays expected cumulative stopping probabilities under various hypothetical references, where the numbers .
For a one-sided design, expected cumulative stopping probabilities at each stage under hypotheses are displayed, where is the alternative reference and are the values specified in the CREF= option.
For a two-sided design, expected cumulative stopping probabilities at each stage under hypotheses and are displayed, where and are the lower and upper alternative references, respectively. Note that for a symmetric two-sided design, only expected cumulative stopping probabilities under hypotheses are derived. The default is CREF= 0 0.5 1.0 1.5.
The following options can be used in the PROC SEQTEST statement to display plots with ODS Graphics. They are listed in alphabetical order.
specifies options that control the details of the plots. The default is PLOTS=TEST. The global plot option ONLY suppresses the default plots and displays only plots specifically requested.
ODS Graphics must be enabled before requesting plots. For example:
ods graphics on; proc seqtest Boundary=Bnd_LDL Parms(Testvar=Trt)=Parms_LDL1 Plots=(test errspend); run; ods graphics off;
For more information about enabling and disabling ODS Graphics, see the section Enabling and Disabling ODS Graphics in Chapter 21, Statistical Graphics Using ODS.
The plot request options include the following.
produces all appropriate plots.
displays a plot of the average sample numbers (expected sample sizes for nonsurvival data or expected number of events for survival data) under various hypothetical references, where the numbers .
For a one-sided design, expected sample numbers under hypotheses are displayed, where is the alternative reference and are the values specified in the CREF= option.
For a two-sided design, expected sample numbers under hypotheses and are displayed, where and are the lower and upper alternative references, respectively. Note that for a symmetric two-sided design, only the average sample numbers under hypotheses are derived. The default is CREF= 0 to 1.5 by 0.01.
displays a plot of conditional powers given the most recently observed statistic under specified hypothetical references, where the numbers . In the SEQTEST procedure, the conditional power is the probability that the test statistic at the final stage would exceed the rejection critical value given the observed statistic.
For a one-sided test, the powers are derived under hypothetical references and , where is the observed statistic, is the alternative reference, and are the values specified in the CREF= option. For a two-sided test, the powers are derived under hypothetical references , , and , where is the lower alternative reference and is the upper alternative reference. The default is CREF= 0 to 1.5 by 0.01.
displays a plot of the error spending for all sequential boundaries in the designs simultaneously. You can display the information level (HSCALE=INFO) or the stage number (HSCALE=STAGE) on the horizontal axis. With HSCALE=INFO, the information fractions are used in the plot. The default is HSCALE=STAGE.
suppresses all plots.
displays a plot of the power curves under various hypothetical references, where the numbers .
For a one-sided design, powers under hypotheses are displayed, where is the alternative reference and are the values specified in the CREF= option.
For a two-sided design, powers under hypotheses and are displayed, where and are the lower and upper alternative references, respectively. Note that for a symmetric two-sided design, only powers under hypotheses are derived. The default is CREF= 0 to 1.5 by 0.01.
displays a plot of repeated confidence intervals. Repeated confidence intervals include both rejection and acceptance confidence intervals.
With the STOP=REJECT or STOP=BOTH option, rejection confidence limits can be derived and the null hypothesis is rejected if the lower rejection confidence limit is greater than or the upper rejection confidence limit is less than .
With the STOP=ACCEPT or STOP=BOTH option, acceptance confidence limits can be derived and the null hypothesis is accepted with alternative hypotheses and if the upper acceptance confidence limit is less than and the lower acceptance confidence limit is greater than .
displays a plot of the sequential boundaries and test variables. Either the information level (HSCALE=INFO) or the sample size (HSCALE=SAMPLESIZE) is displayed on the horizontal axis. The HSCALE=SAMPLESIZE option is applicable only if the sample size information is available in both the input BOUNDARY= data set and input DATA= data set. The stage number for each stage is displayed inside the plot. The default is HSCALE=INFO.