Turning an online experiment into an offline one

Comments

• sebastiaan

  August 2022

  Hi @reversemoonwalk ,

  The main and possible only thing that prevents you from running this experiment on the desktop is that the inline_javascript items use some ECMA 6 syntax, which a recent version of JavaScript that OpenSesame doesn't support yet when running on the desktop. At least the generateRandomArr() function in the set_task_order item would need to be rewritten so that it's compatible with ECMA 5.1 syntax.

  See also:

  • https://osdoc.cogsci.nl/3.3/manual/javascript/about/#versions-of-javascript-ecma-51-and-ecma-6

  — Sebastiaan

  
  
  Check out SigmundAI.eu for our OpenSesame AI assistant!
  
• Fab

  August 2022 edited August 2022

  Hi @reversemoonwalk,

  I'm pressed for time at the moment and so I haven't tried the whole task, but here's some code you can use to replace some of the existing bits in the set_task_order object:

  
  

  / picks 4 digits between 1 and 5 and stores them in an array
  
  var numbers = [1,2,3,4];
  
  function shuffleArray(array) {
  for (var i = array.length - 1; i > 0; i--) {
  var j = Math.floor(Math.random() * (i + 1));
  var temp = array[i];
  array[i] = array[j];
  array[j] = temp;
  }
  return array;
  }
  
  vars.listsarray = shuffleArray(numbers).join('');
  console.log ("listsarray: "+vars.listsarray)
  

  That code is ES5 compatible and will run when you execute the task in Open Sesame.

  The code basically shuffles an array containing the values 1, 2, 3, 4, so that each digit appears once. I seem to remember that this is what you wanted to achieve a random order of presentation of blocks/conditions.

  I haven't had the time to try the whole task to check the rest runs smoothly, so do check carefully.

  Here's my modified version:

  onlinexp Fab.osexp

  Hope this helps,

  Fabrice.

  
• reversemoonwalk

  August 2022

  Thank you @sebastiaan ! Now that I know that is different for local experiments, I will create future experiments accordingly.

  
  

  Thank you @Fab ! The experiment seems to work with this new code; I will test the whole experiment and let you know. You are great!
• reversemoonwalk

  October 2022

  Hello again @ Fab!

  Your version of the experiment gives the following error after the practice part.

  

  The error might be related to "increment_taskcounter" code. Any suggestions?

  
  

  Thanks in advance!

  onlinexp Fab.osexp

  
  
• reversemoonwalk

  October 2022

  @Fab I have tried several things but failed to make the experiment work. Since the error says "listundefined.csv", I think it might be related to the code you created to randomize list numbers in the set_task_order part.
• Fab

  October 2022

  Hi @reversemoonwalk,

  Sorry for the delay in getting back to you. I only just saw your more recent message (the handle in the first one didn't get picked up -you used "@ Fab" with a space in between- and so I did not receive notification).

  I had a look at the task and it looks as if the problem is that the task doesn't seem to treat the listarray variable as an array. This is likely to be due to the way it is built (after the change of method we introduced to make the task compatible with ECMA 5.1).

  So the following line...

  vars.task_list="list"+vars.listsarray[vars.taskcounter-1]+".csv"
  

  ... fails because the use of the [] characters to access a specific column withing an array does not work(because the task does not consider the listarray variable to be an array to start with).

  After investigating it and trying different things, I found a way around it that allows you to keep the Javascript generating the random order of list in a way that will work with ECMA 5.1.

  I modified bits of the code within the increment_taskcounter object to set up a temporary variable (temparray) that contains a string version of the listarray variable plus an empty character. It then becomes possible to use Javascript's Slice function to isolate a single character in that string, which I do using the taskcounter variable).

  slice() extracts a part of a string and returns the extracted part in a new string. The method takes 2 parameters: start position, and end position (end not included). hence, for example,. slice(0,1) extracts the 1st character of a string, while slice(1,2) extracts the second, etc.

  Using this method, it is now possible to isolate the digit from the listarray that correspond to the taskcounter and use it to specify which csv file should be used.

  Here's the new code with some console.log events just to keep monitoring what's going on in the console.

  console.log ("Test phase starts here ________________________________")
  vars.taskcounter+=1
  vars.task_info=vars.task[vars.taskcounter-1]
  
  // allocate csv files to the current task, basd on previously randomy picked list
  console.log("List array:"+vars.listsarray)
  
  console.log ("Sets up temporary string containing the list")
  let temparray = vars.listsarray.toString()+" "
  vars.temp=temparray.slice(vars.taskcounter-1,vars.taskcounter)
  vars.task_list="list"+vars.temp+".csv"
  console.log ("List loaded up for this condition is: "+vars.task_list)
  console.log ("taskcounter: "+vars.taskcounter)
  

  I tried it out up until the second test block but not further. Incidentally, I notice that your list1.csv file contained an error. You had a column header reading "q" instead of "targetldt_tur". I corrected it and uploaded the correct list.csv to the file pool.

  I attach my modified version. Hopefully it'll work. Just make sure to run the whole task to the end to check it ruins correctly to the end.

  onlinexp Fab v2.osexp

  Good luck!

  Fabrice.

  
• reversemoonwalk

  October 2022

  Thanks again @Fab ! I cannot thank you enough. Along with your most beneficial help, you are also helping me to understand what went wrong. I have been testing the new experiment so far, but there is one little problem. Correct answers should be indicated with green lines but I see red lines for some correct answers. When I checked the log file, I saw that some of the correct answers were marked wrong. Can that be related to the code we included for determining correct answers (define_correctresponse)?

  if (vars.category=="word"){
  vars.correct_response=vars.correctword
  } else {
  vars.correct_response=vars.correctnonword
  } 
  

  This is what I did to determine the type of category for each task:

  if (vars.task_info=="Türkçe Sözcüksel Karar Görevi" && vars.subject_nr%2==0){
  vars.prime= "[primeldt1_en]"
  vars.target= "[targetldt_tur]"
  vars.category= "[category_ldt]"
  }
  
  if (vars.task_info=="Türkçe Sözcüksel Karar Görevi" && vars.subject_nr%2==1){
  vars.prime= "[primeldt2_en]"
  vars.target= "[targetldt_tur]"
  vars.category= "[category_ldt]"
  } 
  

  The code above was used in increment_taskcounter to determine the target and prime types for each task and alternate two prime conditions. These seem to work fine.

  
  

  Also, I added 'vars.category= "[category_ldt]"'' part to set the correct answers for that specific list. In my lists, this columns refer to the values "word" and "nonword". But the first code I wrote here seems to fail to differentiate correct answers from wrong ones.

  

  
  
• Fab

  October 2022

  Hi @reversemoonwalk,

  Took me a little while to work out what's going on there (I first tried to work out whether the issue was specific to a particular condition or csv file, etc.), but I think I found the issue.

  I started by adding some code to output to the console what the category and correct responses were, and I then realized what the task scored a correct was based not on what was happening on the current trial but on the previous one. When I started to respond based on what the correct response was on the previous trial, then the red vs green feedback indicated my response was correct every time. Once I realized that, it became easier to spot the problem.

  Basically, the code defining the correct answer is correct but the problem is that it was executed at the wrong time. It turns out that when a trial runs, the correct response variable is set when the sequence is prepared, not when it is run. That means that if you run the code when the trial runs, it will store the correct response but that will only be used the next time the trial sequence is prepared (i.e., on the next trial).

  So, the fix is to make sure that the code defining the correct response is executed when the trial sequence is being prepared, which is done by placing that code in the "prepare" section of the inline_javascript object like this:

  

  
  

  I think this solves the issue.

  onlinexp Fab v2.1.osexp

  Best,

  Fabrice.