Jatos, experiment won't reload because of "finished study", data not saved

Comments

• kri

  April 26 edited May 16

  Hi Giacomo,

  I took some time to answer your long question :)

  Maybe there is a general misunderstanding of how JATOS handles study runs: First you have a study link (e.g. https://example.com/publix/OZTF4BqnlWd), in your case it is of type general multiple and that means you can use this link as often as you want. Then, each time you use this link, JATOS starts a new study run (with a new study result: each line in your result page represents a study result) and redirects the participants to a different type of URL (they are much longer, e.g. https://example.com/publix/4c93ba3f-eda7-45e7-9db0-200e370dc196/15287b49-d9e1-4d20-85c8-a412c9ac2075/start). Often a JATOS study consists of multiple components - then each component has its own component run URL. Those study run URLs are not "general multiple" any more, that means, they have the normal constraints of a study run, like the other study link types (personal single, personal multiple, ...). E.g. if your component is not allowed to reload, a reload in the browser will cause JATOS to end the study with a state FAIL. Or if your study is linear, you cannot loop back to a previous component.

  In your case you seem to have forbidden the reloading of a component (which is often what you want) and your participant reloaded the page (for whatever reason, e.g. errors in the experiment, network problems). Then the study run got finished with a FAIL. Each consecutive reload just returned an error page from JATOS. Then your participant used the original study link again (a general multiple) and started a new study run with a new study result. That worked.

  And to answer some of your questions:

  The first is that sometimes the experiment won't reload because "the study is finished". Specifically, the message is "It's not allowed to reload this component (ID: 21961). Study (ID: 14167) is finished."

  Your component is not allowed to be reloaded. Once it does, it ends with FAIL.

  Note that the two IDs are always the same (even when using different browsers, computers, and networks).

  The IDs belong to the general multiple study link.

  Also, I don't get why the problem then seems to go away by itself (e.g., if I try again at a later time, the study loads with no issues).

  I'm sure the person, at a later time, uses the original general multiple study link again - while before it just pressed reload in the browser.

  I'm not sure I actually helped you with your issues, but I hope I could help you understanding it.

  Best,

  Kristian

  
  

  Edit: Spelling
• giacomospinelli

  April 28

  Hi Kristian,

  thank you very much for your reply. You did help, thank you, and I am now beginning to see what might have happened. All in all, I think it would be safer to use personal, single-use links in order to have a bit more control on what the participant does. There's still the question of why the data for the run the participant did finish were lost even if the participant reached the end of study component. More specifically, I'd like to understand if there's something I can do to moving forward to prevent that from happening again (it's hard to get participants for this study and I wouldn't want to lose too much data). Thanks again for your help!

  
  

  Best,

  Giacomo
• kri

  April 29

  There's still the question of why the data for the run the participant did finish were lost even if the participant reached the end of study component. More specifically, I'd like to understand if there's something I can do to moving forward to prevent that from happening again (it's hard to get participants for this study and I wouldn't want to lose too much data).

  First, JATOS goes to quite some length to ensure that the sent result data reach the JATOS server. Importantly, it has a background worker (called httpLoop) running in the browser of your participants that handles all sending of result data (and other stuff) to the JATOS server. This httpLoop ensures that the order of the result data is kept (in case you sent data several times: 1 - 2 - 3 stays 1 - 2 - 3), then it automatically retries the sending if, for whatever reason, it failed (it retries up to 5 times). And finally it ensures that the study only ends or goes to the next component when all data are sent. So, it should never happen if you use JATOS (and jatos.js) correctly and there is no Internet issue that a study FINISHED and data get lost. That said, people can use jatos.js' control flow functions or functions to send data/files wrong. Can you post here the part of your code where you send your data and finish your study / go the next component? I can have a look.

  Best,

  K.
• giacomospinelli

  May 11

  Hi again Kristian,

  sorry for the hiatus and thanks again for your help. I honestly didn't look at those pages because I just used, with minimal customization, the code reported on this page for integrating jspsych and jatos. Specifically, in the <head> element I included the following:

  <script src="jatos.js"></script>

  Then at the very beginning of the script element, the following:

  var jsPsych = initJsPsych({

     use_webaudio: false,

     on_finish: () => jatos.endStudy(jsPsych.data.get().json())

    });

  And finally at the very end of the script element (after all of the jspsych timeline variable has been defined), the following:

    jatos.onLoad(() => {

     jsPsych.run(timeline);

    });

  What do you think? Thanks!

  
  

  Best,

  Giacomo
• kri

  May 16

  Hi Giacomo,

  Sorry, for taking some time to answer. Your code looks good. It's the standard jatos-wrapping around a jsPsych experiment.

  Another thing that comes to my mind: what is the size of your result data? There is a quota of 5 MB per result data sending and if you exceed this the JATOS server will not store the data in its database.

  Best,

  Kristian
• giacomospinelli

  May 16

  Hi Kristian,

  you mean the text file (the one you get from data only, plain text)? That's around 120 KB per participant.

  
  

  Best,

  Giacomo
• kri

  May 16

  120kB is far from problematic. So, I am out of ideas, apart from problems with the participant's network or computer, or your study's code.

  K.
• giacomospinelli

  June 14

  Hi Kristian,

  that's OK, thanks for all the help!

  
  

  Best,

  Giacomo