Dear Stephen,

 

If you look at the source of VBOOTSTRAP or VCRITICAL you will get the method of how to do this

(use Help | Procedure Source and select VBOOTSTRAP).

You need to iterate some extra times and stop when you have the required number

of successful simulations. You turn off stopping for faults and then check the exit codes

and any faults set by Genstat. The explanation of the key steps is here:

 

SCALAR i,nonconverg; VALUE=0 "Number of successes and failures"

CALC   maxnboot = NBOOT + NRETRIES

FOR [NTIMES=maxnboot; INDEX=k]

  SET [DIAGNOSTIC=*; FAULT=*] "Turn off stopping for faults and clear any fault code"

  FIT ... "Fit your model"

  RKEEP EXIT=lastexit "Get exit code"

  SET [DIAGNOSTIC=fault] "Turn on stopping for faults back on"

  "Check for any error that has not set EXIT parameter to nonzero value"

  GET  [FAULT=Fault]

  IF  Fault .nes. '    ' .AND. lastexit == 0

    CALC lastexit = -3

  ENDIF

  IF lastexit == 0

     CALC i = i + 1

     CALC Sim$[i] = ... "Save results from this simulation"

  ELSE

     CALC nonconverg = nonconverg + 1

  ENDIF

  EXIT [CONTROL=for] i == NBOOT "Stop when we have enough successes"

ENDFOR

 

There is also the SET [ACTIONAFTERFAULT=continue] option which will give

the fault messages and continue the for loop. The advantage of using the

SET [DIAGNOSTIC=*] is that the fault messages are suppressed.

 

 

I am running a simulation in which I am carrying out a logistic regression for the proportional odds model. Every now and then it stops because it predicts a negative probability. I should like the program to simply ignore the result and proceed with the next simulation. Is there an easy way to do this?

Stephen Senn

 

Please note new email

Stephen Senn

 

[log in to unmask]

Tel +44 (0)7811166929

 

 

To unsubscribe from the GENSTAT list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=GENSTAT&A=1