S+SeqTrial Known Problems and Solutions

This page lists known problems in SeqTrial 1.0, release 1, and contains a link to S-PLUS code that fixes some of the problems.

Known Problems

1. If you are using an older version of Adobe Acrobat that does not allow spaces in folder names, you may not be able to access the S+SeqTrial User's Guide from within S-PLUS. Three solutions are
   
   1. update to a newer version of Acrobat,
   2. open your module folder (e.g. C:\Program Files\sp2000\module\seqtrial), and move the sequser.pdf document to another location to open it, or
   3. start Acrobat first, then open the document as e.g. C:\Progra~1\sp2000\module\seqtrial\sequser.pdf)
2. The module requires S+2000, not S+4.5. With S+4.5 the graphical user interface is not available, and some command line calls will fail.
3. seqMonitor fails to converge if the alternative hypothesis is less than the null hypothesis.
   A patch is available below.
4. help(seqDesign) indicates that the output may be used by the seqHypothesis function. There is no seqHypothesis function. A description of the seqHypothesis object is available by doing help(seqHypothesis.object).
5. 'summary(a two-sided design)' has default values for theta which are all on the same side of the null hypothesis, and contain duplicates. From the Graphical User Interface (GUI) this occurs when requesting "Summary Tables" on the "Results" tab.
   A patch is available below.
6. There is currently no graphical interface for monitoring capabilities. Monitoring is performed by the `seqMonitor' function from the command line. To obtain help on this function, do `help(seqMonitor)' from the command line, or start SeqTrial Help, choose the "Find" tab, type monitor, and click on "Group Sequential Monitoring" at the bottom.
7. There is currently no graphical interface for `seqPHSampleSize', the function which estimates the number of patients to accrue in order to obtain a required number of events for a survival trial. Use this function from the command line.
8. When creating designs with constraints, GUI rollback fails to save the constraint matrices (i.e. if you switch back to a previous design using the small arrows at the bottom of the GUI window). You may enter them again. Or, if the design was saved by name, then use "Update Design" under the SeqTrial menu.
9. A number of the links in the help files do not work. In most cases the corresponding functions (options, panel.superpose, par, points, polygon, trellis.args, trellis.par.set) are part of S+2000. You can get help for these by doing e.g. help(options) from the command line. Two cases are typos; instead of seqPlotAsn, do help(seqPlotASN) from the command line, and for panel.stop.prob, do help(panel.seqPlotStopProb) Finally, these two help files do not exist: update.seqDesign (see update instead), and seqMonitor.object.
10. When monitoring (using seqMonitor) a hazard ratio design, one or more messages are printed indicating that: "Argument variance ignored in proportional hazards probability model". They arise due to the way that seqMonitor calls seqDesign. Just ignore them.
11. In the hazard model case, when a log transformation is not used, seqMonitor fails to convert the log-hazard-ratio calculated from the data to a hazard-ratio.

This patches file contains fixes for some problems noted above. Download it, and load it into S-PLUS using the directions contained in the file.

Please report any additional problems, or suggestions for improvements, to support@solutionmetrics.com.au

Last updated 19 March 2001.