[solved] Staircase

Comments

• Wouter

  November 2012 edited November 2012

  Hi boo,

  Sebastiaan contacted me about this.

  as for the original post, I was able to figure it out:
  http://www.filedropper.com/ioraccletters1opensesametar

  Bare with me; it was the first thing I ever did in python (and thereby also in opensesame). The whole thing is a bit bloated, and it's full of eyelink-stuff that you probably won't need; but I hope you can still use it as an inspiration

  The task was a slight variation of the posner cueing paradigm, where participants were to indicate whether an M or N was presented. The presentation time of the letter is what I varied using PEST, as outlined here:
  http://www.hitl.washington.edu/publications/r-98-11/node54.html
  (I'm not sure what you meant by "up to 8 different display timings", but maybe using PEST provides an alternative?)

  My expt looks as follows;

  • An introduction with some familiarizing to the displays

  • A practice loop where the pres. time is not varied yet.

  • A staircase loop sc_loop where PEST is first run; either until it has converged or until It has lasted 'long enough'

  • An experiment during which PEST runs in the background

  The staircase loop is what you're probably most interested in. What I did was:

  1. Have a staircase sequence, and within this sequence, place a loop that runs over N (for me N was 9) trials of randomly chosen conditions. (setting N close to all possible combinations makes your staircase fully balanced, but would in my case imply a tediously long staircase phase)
  2. Within this loop, the script DoPEST checks the accuracy every N trials and changes the critical variable(s) (here 'prestime') accordingly. As PEST requires, it also keeps track of these changes
  3. in DoPEST it is also determined whether 'continue_staircase' should be set to False
  4. at the end of the sc_sequence, a script 'check_rerun_sc_loop' eventually does the magic: it will rerun the sc_block_loop (i.e. N trials, and the PEST algorithm) in the way discussed in the previous thread:
     

     
     
     while self.get('continue_staircase') == True:
     
         self.experiment.items['sc_block_loop'].prepare()
     
         self.experiment.items['sc_block_loop'].run()
     
     

  ... thus breaking from it when 'continue_staircase' becomes false.

  That's practically all! For you, looking at the sc_loop structure seems most useful. I hope you can work it out this way!

  It is a quite powerfool way to manipulate experiment flow, if you want more than simple randomization, and I use it a lot currently. One word of warning though: using while loops is always tricky: make sure you will at least ever break from your staircase procedure

  good luck!

  Wouter
• booradley

  November 2012 edited 12:32PM

  Hi there Wouter,

  Thanks very much for getting back to me!

  I've downloaded your OpenSesame file but unfortunately get the following error:

    Error: Failed to open 'C:/Documents and Settings/psundere/Desktop/Staircase        /IOR_acc_letters1.opensesame.tar.gz'
    Description: not a gzip file
  

  I was wondering what version of OpenSesame you used as I'm still operating on 0.26?

  If you use 0.27 this may be why I cannot open your file.

  Boo.
• Wouter

  November 2012 edited 12:32PM

  huh...remarkable...I'm under 0.26 as well, so that can't be it.
  here's the script by itself; i.e. not as a tar.gz but simply a .opensesame; does that work?

  http://www.filedropper.com/ioraccletters1
• booradley

  November 2012 edited 12:32PM

  Hi,

  That opens up just fine now

  I'll go ahead and have a play around with it for the rest of the day and report back how I get on.

  Thanks Wouter, you're a star

  Boo.
• booradley

  December 2012 edited 12:32PM

  Hi Wouter,

  I ended up doing something very different from your code and it also worked. I'll post it up here with detailed descriptions on what is happening soon for any other user that wants to do something similar.

  If mods could leave the thread open for this purpose I'd appreciate it.

  Thanks all,

  Boo.
• booradley

  February 2013 edited February 2013

  Hi all,

  Sorry for taking so long to add this. Xmas break and crazy work since has distracted me somewhat.

  The following is the staircase code I used for my experimental design in OS.

  It is by no means the most efficient code (it certainly could be cleaned up).

  It is a 6 block sequence where at the end of each block if accuracy falls on, above or below 75% the duration of display is adjust. For example, if below 75% the display duration is increased by 20ms.

  After 3 blocks instead of 20 the changes are one frame of 10ms.

  Here are shots of the OS stream of events:

  file:///F:/open_forum/pic.bmp

  The following is the code used in the inline script called 'acc_script':

  print "START OF INLINE SCRIPT"

  dur1 = self.experiment.get("dur")
  acc = self.experiment.get("accuracy")
  loop_count = self.experiment.get("count_inline_loop")

  correct_count1 = self.experiment.get("correct_count")

  if ((loop_count+1) <= 3):
  print "FIRST IF < 3 LOOP COUNT"
  print loop_count
  if (correct_count1 > 10 ):
  print "correct_count1 > 10", correct_count1
  self.experiment.items['TOJ_Seq'].prepare()
  self.experiment.unset("dur")
  self.experiment.set("dur", dur1 - 20)
  self.experiment.unset("accuracy")
  self.experiment.set("accuracy", 0)
  self.experiment.unset("correct_count")
  self.experiment.set("correct_count", 0)
  self.experiment.items['TOJ_Seq'].run()

  elif (correct_count1 < 8):
      print "correct_count1 < 8", correct_count1
      self.experiment.items['TOJ_Seq'].prepare()
      self.experiment.unset("dur")
      self.experiment.set("dur", dur1 + 20)
      self.experiment.unset("accuracy")
      self.experiment.set("accuracy", 0)
      self.experiment.unset("correct_count")
      self.experiment.set("correct_count", 0)
      self.experiment.items['TOJ_Seq'].run()
  
  elif (correct_count1 == 8):
      print "correct_count1 = 8", correct_count1
      self.experiment.items['TOJ_Seq'].prepare()
      self.experiment.unset("dur")
      self.experiment.set("dur", dur1 + 10)
      self.experiment.unset("accuracy")
      self.experiment.set("accuracy", 0)
      self.experiment.unset("correct_count")
      self.experiment.set("correct_count", 0)
      self.experiment.items['TOJ_Seq'].run()
  
  elif (correct_count1 == 10):
      print "correct_count1 = 10", correct_count1
      self.experiment.items['TOJ_Seq'].prepare()
      self.experiment.unset("dur")
      self.experiment.set("dur", dur1 - 10)
      self.experiment.unset("accuracy")
      self.experiment.set("accuracy", 0)
      self.experiment.unset("accuracy")
      self.experiment.set("accuracy", 0)
      self.experiment.items['TOJ_Seq'].run()
  
  else:
      print "Keep timing as is", dur1
      self.experiment.items['TOJ_Seq'].prepare()
      self.experiment.unset("correct_count")
      self.experiment.set("correct_count", 0)
      self.experiment.unset("accuracy")
      self.experiment.set("accuracy", 0)
      self.experiment.items['TOJ_Seq'].run()
  

  elif ((loop_count+1) >= 4) and ((loop_count+1) <= 6):
  print "ELIF > 3 LOOP COUNT"

  if ((loop_count+1) < 6) and (correct_count1 > 10):
      print "correct_count1 > 10", correct_count1
      self.experiment.items['TOJ_Seq'].prepare()
      self.experiment.unset("dur")
      self.experiment.set("dur", dur1 - 10)
      self.experiment.unset("accuracy")
      self.experiment.set("accuracy", 0)
      self.experiment.unset("correct_count")
      self.experiment.set("correct_count", 0)
      self.experiment.items['TOJ_Seq'].run()
  
  elif ((loop_count+1) < 6) and (correct_count1 <= 7):
      print "corerct_count1 = 7", correct_count1
      self.experiment.items['TOJ_Seq'].prepare()
      self.experiment.unset("dur")
      self.experiment.set("dur", dur1 + 10)
      self.experiment.unset("accuracy")
      self.experiment.set("accuracy", 0)
      self.experiment.unset("correct_count")
      self.experiment.set("correct_count", 0)
      self.experiment.items['TOJ_Seq'].run()
  
  elif ((loop_count+1) < 6) and (correct_count1 >=8) and (correct_count1 <= 10):
      print "correct_count1 is 8-10", correct_count1
      self.experiment.items['TOJ_Seq'].prepare()
      self.experiment.unset("accuracy")
      self.experiment.set("accuracy", 0)
      self.experiment.unset("correct_count")
      self.experiment.set("correct_count", 0)
      self.experiment.items['TOJ_Seq'].run()
  
  elif ((loop_count+1) == 6) and (correct_count1 > 10):
      print loop_count
      print "END OF LAST BLOCK GO TO EXPERIMENT"
      print "correct_count1 > 10", correct_count1
      self.experiment.items['Experimental_Loop'].prepare()
      self.experiment.unset("dur")
      self.experiment.set("dur", dur1 - 10)
      self.experiment.unset("accuracy")
      self.experiment.set("accuracy", 0)
      self.experiment.unset("correct_count")
      self.experiment.set("correct_count", 0)
      self.experiment.items['Experimental_Loop'].run()
  
  
  elif ((loop_count+1) == 6) and (correct_count1 <= 7):
      print "corerct_count1 = 7", correct_count1
      self.experiment.items['Experimental_Loop'].prepare()
      self.experiment.unset("dur")
      self.experiment.set("dur", dur1 + 10)
      self.experiment.unset("accuracy")
      self.experiment.set("accuracy", 0)
      self.experiment.unset("correct_count")
      self.experiment.set("correct_count", 0)
      self.experiment.items['Experimental_Loop'].run()
  
  elif ((loop_count+1) == 6) and (correct_count1 >=8) and (correct_count1 <= 10):
      print "correct_count1 is 8-10", correct_count1
      self.experiment.items['Experimental_Loop'].prepare()
      self.experiment.unset("accuracy")
      self.experiment.set("accuracy", 0)
      self.experiment.unset("correct_count")
      self.experiment.set("correct_count", 0)
      self.experiment.items['Experimental_Loop'].run()
  
  
  else:
      print "SOMETHING WRONG in > 3 loop count"
  

  else:
  print "SOMETHING WENT WRONG!!!!"

  The variable 'dur' was defined in the 'set_variables' loop.

  I hope someone can find a use for this basic code or at least it may point you in the right direction.