Export confirmation codes, add labels to results about jatos HOT 18 CLOSED

It should be possible. Let me give it a try.

On Tue, Nov 18, 2014 at 2:57 PM, elisafilevich [email protected]
wrote:

Assigned #40 #40 to
@kristian-lange https://github.com/kristian-lange.

—
Reply to this email directly or view it on GitHub
#40 (comment).

from jatos.

Do you still think this is a good idea? I actually think we should wait for more feedback from the other users before implementing this.

from jatos.

I do think that it would be a good idea. But I already wrote a way around it in my matlab analysis so I don'T need it right now. Let's leave it as an option and wait until we get feedback, yes

from jatos.

I'm analyzing data again. This would definitely help. The question is, what if people store csv and not JSON as output data? Let's discuss any alternatives

from jatos.

Another alternative is: have (yet another) button, or export an additional file with suffix '_metadata', that contains the metadata of each one of the exported results in JSON format. I know it's suboptimal to export two files, but at least it gets rid of the problems that might arise if people are exporting their results in csv format.
I really need this. It's useful to keep track of whether the experiment was really finished (although if we look at the table behaviour we might not need this) and how long it took.

from jatos.

Has there been any update on this? I would like to provide performance dependent payoff in mturk and for this I would need to match the mturk worker id, which is available in Jatos, with with my results. I could match them based on the worker id which is present in both data sets, but I do not seem to be able to export the list of mturk worker ids. At least, I do not seem able to export the metadata.

from jatos.

I could add a button in JATOS result page that exports the result data together with the metadata. And here I could need some input from you as a user: I would use JSON format for the whole results (result data + metadata). So the whole exported file is an JSON array with each result an element in this array. Each result has all the meta data and a field 'resultData' which contains the result data using JSON encoding.

It would look similar to this:

[
{"resultId": 1, "startTime": 12345678, "workerId": 1, ... , "resultData": "all data in JSON encoding"},
{"resultId": 2, ...
]

Does this suits your needs?

Btw. I could also make the MTurk worker ID available in jatos.js / jatos.urlQueryParameters. From there one can do with them as pleased and even put them in the result data where they will be stored in JATOS database. Would this help?

from jatos.

I use lab.js for programming my experiment and then export to JATOS, which makes the data available as a JSON. So what you suggest would probably work quite well.

However, only having the metadata in a separate JSON file (i.e., without resultData) would work as well as matching via resultID is also quite trivial. I do not think exporting two files would be such a hassle. However, if your initial suggestion is easier for you to implement, please do that. Either way is fine.

Finally, I read the statement on privacy and figured that is the reason it is not possible. However, allowing performance related payment is something I think quite a lot of users would want (I for sure would be super happy). So maybe it is a great way to allow this somewhat easier via the API. In the end, we users should be aware that we are ultimately responsible for ensuring the data of our participants are treated in an ethical manner and identifying information, such as the MTurk IDs, are not used in an inappropriate or illegal manner or even posted on the web.

from jatos.

from jatos.

Thanks, that sound great. And this is not super urgent. I have a project planned where we want to use it, but the PhD student is still in the programming of the experiment stage.

And does your corrected statement mean that the mturk ID is already available via the API?

from jatos.

No the MTurk worker ID is not available via jatos.js yet. I'll add it in the next release.

from jatos.

There are other URL query parameters next to the MTurk workerId coming from MTurk:
assigmentId, hitId, turkSubmitTo. Do you think they could be of any use for a study? If not I would rather not add them to keep data as minimal as possible.

from jatos.

I do not think they would be of any use. From my perspective there is no need to add them. Looking forward to the next release!

from jatos.

Hej everybody, just a quick note regarding a possible workaround for this specific case -- lab.js also captures query parameters (at least it should), so they should be available in the main data (all URL parameters should be in the column url). That being said, managing this information through JATOS seems like a much better idea!

Added upon reflection: I see the privacy angle, and should probably rethink our library's behavior here. Would be happy to discuss this!

from jatos.

@FelixHenninger Unfortunately, the MTurkId is not part of the URL participants click on. I believe it is only part of the return URL (or something like that). At least, it is clearly not part of the data. I only get the subject ID assigned by JATOS. For example, the data of one participants begins with:
[{"url":{"srid":"50"},"meta" ...

from jatos.

from jatos.

This issue is getting confusing - it spread out into multiple issues :/

Good news: MTurk worker ID is implemented: jatos.urlQueryParameters.workerId is part of next release

Question regarding result metadata: I would implement it like you suggested without the actual result data - just metadata. And then JSON is actually overkill. Since it's already in table format I'd prefer CSV format. Do you agree?

from jatos.

Yes, I agree. csv for tabular data is perfect.

And sorry for the confusion.

from jatos.