# Difference between revisions of "Participant Counter"

Imagine you are building an online experiment that will use an external panel system. The participants will reach the experiment via a link supplied by the panel system The panel system will give you a participant identifier like this: ae56b232. Now you want to counterbalance your conditions. Normally you would do this by making a list of conditions:

participant condition a condition b
0 Red Horizontal
1 Green Horizontal
2 Red Vertical
3 Green Vertical

Now you would normally use the participant number to determine the condition: participant 0 get red, horizontal, participant 1 get green horizontal, participant 0 get red, vertical, participant 2 get green vertical en participant 2 gets red, horizontal again. (in real life the number of conditional can be much larger of course) This nice system is ruined by the fact that we have no consecutive participant numbers. Using the random participant numbersfrom the panel system may work if the number of participant is very large and the number of conditions is very small, but otherwise our counterbalancing is ruined.

Use the participant counter to solve this. Just give the participant counter a nice unique name for our experiment and each time you call it, it will give you a new counter value. It has some additional features as well.

# just count

# {"experiment":"1","counter":"0","timeout":""} # first counter value
# {"experiment":"1","counter":"1","timeout":""} # a new counter value
# {"experiment":"1","counter":"2","timeout":""} # and yet another counter value

# count with timeout

# {"experiment":"1","counter":"3","timeout":"10"} # ppn is released after ten seconds
sleep 15
# {"experiment":"1","counter":"3","timeout":"10"} # we get the same counter value again

## finalizing counter at the end of an experiment

### experiment finished in time, finalizing successful

# {"experiment":"1","counter":"3","timeout":"10"}
# {"experiment":"1","counter":"3","timeout":""} # counter value is finalized, no other participant will receive this value in the future

### experiment takes too long and finalizing fails

curl 'https://www.socsci.ru.nl/wilberth/nocms/radcloud/participant_counter.php?experiment=1&timeout=10' # get counter value
# {"experiment":"1","counter":"3","timeout":"10"}
sleep 15 # wait too long
curl 'https://www.socsci.ru.nl/wilberth/nocms/radcloud/participant_counter.php?experiment=1&counter=3' # attempt to finalize counter
# {"experiment":"1","counter":"","timeout":""} # attempt fails, your were late (counter is not returned)

# Automatic forwarding to experiment

You need not read the result as JSON. In order to get a 302 temprary redirect to https://www.nu.nl/?experiment=1&counter=8&timeout=10 use:

All the other features work in combination with the url parameter as well.

# Adding a participant identifier from the panel system

You can add the participant value that you received from the panel system. This way, if the participant returns to your page (or for instances presses F5) he will receive the same counter value and hence the same version of the experiment.

# {"experiment":"1","counter":"0","timeout":"","participant:"john"} # first counter value
# {"experiment":"1","counter":"0","timeout":"","participant:"john"} # The same counter value that john received last time.

If you combine a participant value with a timeout, the timeout will be reset to the new value, giving the participant more time to complete the experiment. Setting timeout on a finalized experiment will not result in an error, but will not reset the timeout value.

# Maximum counter value

This feature has not yet been implemented

# Set of counter values

This feature has not yet been implemented

# Random order of counter values

This feature has not yet been implemented