presleyp / speriment Goto Github PK

Make them not count for decision about which condition/counterbalance to use next. Could work for exclusion criteria that are based on one participant, ie ReactionTime but not whether they're an outlier.

Add feature to distribute resources or trials across blocks randomly per participant.

If it makes it hard to import into R

Actually, not sure this is necessary, but it might be a useful shortcut for users.

This is the most critical thing missing for my alternations experiment. I want to match pictures to words differently across participants, but once a picture-word association has been chosen for a participant, I need the related word to get the related picture so that singulars and plurals share semantics as expected.

I'm not sure the best way to do this. A general way would be to add something that knows how to set per-participant ground truth and then train on that. But that is probably too general; constraints and training styles probably vary too much.

Another way would be an object like SampleFrom, something like MapFrom. Whereas SampleFrom takes a string from a list, MapFrom would take a string from a dictionary. The key used to access the string would be a page id, so it would mean "get me the corresponding value to the value page 1 used." The experimenter could write in the other half of the correspondence, by putting in correlated text in these two pages for instance.

So I could have two banks of pictures. The singular bank would be a list. The plural bank would be a dictionary from singulars to plurals. When a string was sampled from the singular bank, the page id it went to would become the key to the plural in the plural bank. When MapFrom is called, it accesses that plural.

It's not terribly pretty but it's a lot prettier than all the other ways I can think of. I'll have to think about more complex versions we might want and ways to make it more flexible. At least this would mean you could map from different pages in the same page, and different aspects of a page.

Maybe options to filter by experiment version or date, or maybe allow a SQL query to be passed in and only return results from that. Just excluding by worker ID isn't very convenient.

Can I write a script to do PsiTurk commands? Take a function written by the user to decide who should get a bonus? Could be reused in the decision to exclude people.

Should only run if pseudorandom is set.

Very strange bug where Qunit insists on a filter, apparently of a certain length. Version 1.14 is fine but I'll want to figure this out for the future.

Unless there's a http(s):// in front of a link, I think (GitHub flavored) markdown treats links as relative. Because of this, the link to the PsiTurk documentation in the README currently points to https://github.com/presleyp/Speriment/blob/master/psiturk.readthedocs.org, rather than https://psiturk.readthedocs.org.

This probably means introducing a new Trial class that contains one or more Pages.

Should be saved somewhere and accessed by speriment-output.

And make them sample-able.

Record when a sound or video is played.

Easier to parse lists in R if they're 'c(1,2,3)' rather than '[1,2,3]'.

So much more needed - specifically I want to remember to check for exchangeable-counterbalance overlap.

OuterBlocks are initializing properly but not displaying things.

Right now, all non-text questions have to have answers for Next to be enabled. It would be better to have all questions require answers uniformly, and give all an argument optional that can be set to allow participants to skip questions.

This would be easy to do but I have to think about how it should work. If InnerBlocks don't have criterion, fine. If they do, does it calculate only over their last iteration?

In release 7 I add resetting of option feedback pages in a training block, but there's still a problem.

Counterbalancing may be for ordering blocks or for choosing blocks. Make the latter possible using RunIf.

http://www.crockford.com/javascript/jsmin.html

• PsiTurk's condition variable has sensical values
• Latin Square unit tests pass
• works in viewpage.html
• With PsiTurk, groups get initialized correctly
• It seems to go wrong in the latin square method, but only when PsiTurk is involved

It turns out to be useful in the category learning experiment.

Possibly replacing mouse interaction entirely.

after writing text

Should be easy, just let feedback be Pages. Good for letting people study image-word association.

So you can use AWS instance or MySQL or something.

Make blocks contingent on whether a participant passed a training block within a certain number of iterations.

It would be nice to be able to look at, for instance, answers to freetext questions by querying the database. But they're buried in the JSON object in the datastring column. A tool that takes the name of the JSON property and creates a column for it would help.

Should be available in Page and Option. SampleFrom should work inside.

Conditional running of blocks won't work in Release 2. It's a one-line fix, will patch soon.

Use PsiTurk's counterbalance variable to inform block exchanging.

PsiTurk's CSS doesn't look right with my long consent forms. I put a fix in speriment.css. If you need it, add <link rel=stylesheet href="/static/lib/node_modules/speriment/css/speriment.css" type="text/css">
to templates/consent.html.

Because they can be sampled. Should be paired with #6.

When setting an ID of one type, check if the same ID has already been made of the other type.

Also check within-type for user-defined IDs.

• commit changes to API so it shows on rawgit
• add docs on using MySQL - for instance, need to start its server before starting PsiTurk server

Only play sound once, play sound on page load, disable selections until sound is done playing, etc.

to save loading time, especially now that I'm moving to preloading

Meant to have it without replacement by default.

Make exchangeable take a list of lists.

Feedback iteration increases when copied block runs for the first time. Only affects recording as far as I can tell.