Suppose I have about 100 people that will answer to a survey and there are 10 versions of the survey -- for example 10 slightly different pictures and questions about how the respondent liked the picture. I can make a hidden question of type 'equation' having just rand(1, 10) as value and foo as name, and the use foo==1, foo==2 etc. as relevance equations for question of type 'text display'.

However the distribution will probably not be very equal-sized, as the number of people taking survey is quite small. Is there a workaround for this?

For example, can I refer to number of people already filled the survey in expression manager and then just compute the remainder when dividing by ten?

Hi,
as long as the quota management of LS is as basic, you have to query the database to see how many respondents already answered the different versions of the survey.
Do an ajax call in the first question; query the database by direct access or by the API.
Then you should use a "least filled" approach.

Do the respondents get an incentive?
If not, it shouldn't be an issue if you just wait until you have ten responses for each version. The slight overquote in some versions you may remove at random.


Joffm

Joffm wrote: Do an ajax call in the first question; query the database by direct access or by the API.

Thanks. I have never used ajax, and only used API from an outside php page. Can you give some example about using API inside the survey?

Jmantysalo wrote: However the distribution will probably not be very equal-sized, as the number of people taking survey is quite small.

If you had the technical feature to do it inside LimeSurvey, you might still have an uneven distribution after qualitycheck. Completed surveys by speedster or with bullshit are mostly removed afterwards. Even with a more sophisticated quota system (which is LimeSurvey lacking) this issue cannot be resolved.

jelo wrote: Completed surveys by speedster or with bullshit are mostly removed afterwards. Even with a more sophisticated quota system (which is LimeSurvey lacking) this issue cannot be resolved.

This is true, of course, but having one problem is better than having two.

Can you give some example about using API inside the survey?

The API cannot be used from within LimeSurvey (yet). You will need to make an AJAX call to a remote PHP file that uses the API.

OK. Then it is probably easier to make a php page that has just counter. Then the survey could have a "question" of type equation calling that page and computing remainder of it's value.

In my opinion :
Use gitlab.com/SondagesPro/ExportAndStats/getStatInSurvey
To do a JS solution (hidding biggest choice).

Got another idea: If I just use {SAVEDID}? Then I don't need an outside php to get a serial number.

Savedid is given to anyone that start the survey. So it does not guarantee that you have x respondents completed if there are x Savedids that end with 1 for example.

It would be roughly the same problem as for the rand(1,10), probably.

Now if you know all 100 respondents and you are sure all will participate and complete, than you could maybe create a token table and randomly assign the different "paths" to the respondents.

holch wrote: Savedid is given to anyone that start the survey. So it does not guarantee that you have x respondents completed if there are x Savedids that end with 1 for example.

It would be roughly the same problem as for the rand(1,10), probably.

I think it will be at least as good as using rand().

Anyways, kind of strange that one can do about everything but can't refer to quota or number of complete responses. Maybe something like getStatInSurvey-plugin will be part of core in the future.

Jmantysalo wrote: Anyways, kind of strange that one can do about everything but can't refer to quota or number of complete responses. Maybe something like getStatInSurvey-plugin will be part of core in the future.

github.com/LimeSurvey/LimeSurvey/pull/13...187a239308679b427506