Quota is often used to get a sample reflecting certain characteristics of a population. These characteristics are often needed throughout the interview to ask certain questions to certain groups.

E.g we have a sample n=1000.

Sex (n=500 female / n=500 male)
Eating Behavior (n=300 Vegetarian / N=700 Non-Vegetarian)

If we have a joint occurrence we might have something like this:
Vegetarian, female 250
Vegetarian, male 250
Non-Vegetarian, female 250
Non-Vegetarian, male 250

Please correct me, if I am wrong:
Currently it isn't possible to decide after how many screener questions the quota should be calculated and only a logic OR is possible when using more than one question. And the quota variables cannot be passed in via an url link.

It would be interesting if you could label a group of screenquestions as a segment and use these segment variable throughout the survey.

Let's say you have a screenquestion multiple option:
Which city would you like to visit?
1.London
2.Rome
3.Paris
4.Berlin
5.Barcelona
6.Amsterdam
7.Stockholm
8.Prague

Your sample is N=400 and you want to get n=50 for every city and
want 8 different exit Urls (since you have a separate survey for every city).

What can be implemented with workaround, what shoud be implemented as a feature? Who else is interested in better quota and skip control and what is already planned or available in LS2?

Hmmm, somehow I don't get the problem. Quota are there to decide whether a participant should proceed or not (screening process). But if you need the characteristics from the screener throughout the questionnaire, you would use conditions, wouldn't you?

So I am not getting the problem.

I mentioned a few problems. Do you see no problem at all?

The part with assigning a group of rules/conditions to a new name (segmentvariable) would be a lot less faulty since you then only need one condition to ask for "If respondent is in segment A then show him this question" instead of copying the conditions. If the definition of the segment is changed you would need to change a lot of conditions.


Let's take my last example. How would you make sure that you get 50 for each city and assigning them to a dedicated city survey?
Each respondent can check 0-8 cities. If 0 he is screened out. If he checks one city and city is > 50 he is screened out. If he checks 2-9 cities he is not screened out as long as one of these cities is <50. The dedicated city survey has to be assigned via quota exit since limesurvey is still missing multiple EndURLs (My submitted idea was marked as implemented, but when it comes to different EndURLs I haven't found anything yes ideas.limesurvey.org/ideatorrent/idea/144/ ).

OK. I get your point with the cities and I also get your point with the "joint screening criterias".

However, from most other applications (confirmit, globalpark, spss) I usually see that what they do is to create a hidden question with the new variable. Seems like they don't have any other solution as well. But those hidden questions should be possible with Limesurvey, right? (never have used so thorough screening criteria and quota), as most of my online surveys do ask the target group per se and we want to see natural distribution.

Thanks for your time. How would the 8 cities example be solved in Confirmit/Globalpark?

You meant something like this:
www.confirmitexperts.com/2011/07/22/rand...controlled-by-quota/

To allocate respondents to different subgroups should be a valuable feature. It prevents that you waste peoples time and keep invitations and field-time at the minimum.

Since you mentioned "natural distribution". The city example is not changing the natural distribution. If you have a respondent which would like to go to every city, it seem wise to me to let him answer the least popular city survey instead overfull-filling the quota of the most popular city. What I see is that some are looping and have no problem asking people 50min and let them do two surveys. That would be unacceptable to me.

No, with "natural distribution" I meant that for most of my surveys I don't screen at all or only very minor. Sorry if that wasn't clear.

I am not an expert on programming in Confirmit or Globalpark. I just have quite a few surveys from those tools through my hands for revision and during this process often the hidden questions are still visible, so you can see how some stuff is solved. Most of the time I am quite under time pressure so that I just ignore those pages and skip them. I'll have a look at it tonight. Maybe I can think of something. Because at the moment this is too heavy stuff...

jelo, have a look at the new expression manager at Limesurvey 2.0: docs.limesurvey.org/Expression+Manager+HowTos

This is a very powerfull tool that should cover of of the cases you describe e.g. you can populate a hidden question with a certain value IF sex=male AND food=meet.

One issue you probably can't cover with this feature is counting how many people have already ticked the Londn/Rome/Paris box. Therefore you need to use quotas but these could then rely on the populated hidden questions as well.

Hi,

You can assign respondents randomly to groups using the workaround i wrote here: docs.limesurvey.org/Workarounds%3A+Quest...nt_wording_and_media

if you then set a hidden question and populate it with data from the token then you can set a quota on it.

some purely javascript versions are available but i cannot find them anymore. This should be similar to the confirmit solution you presented.

However, i I understand correctly you want (1) to control dynamically the quotas moving respondents from quotas that are full to quotas that are not full and (2) favour the least "popular" quota. This is a tough cookie in whatever software.

You can do as follows:

- offer the possibility to chose 3 cities (or whatever number). Use a multiple drop-down.
- using INSERTANS placeholders to spell out the city create 3 batches of questions, one per city.
- if a city becomes full, delete the option from the initial question. You should be able to do this since they are codes and not fields in the database. Not ideal but....

On a separate note, you *can* create multiple exit links by inserting placeholders. Still, I am not sure how you can use them for your needs.

atiut wrote:
Thanks for your time. You're right. But I don't use tokens. So the workaround in the first paragraph cannot be applied.atiut wrote: You just limit the choice to three. In my case the respondent should tick as much as he liked. If I ask for ranking or a e.g 3 items limits I could just use a random number.

You're idea with removing the item is nice and might work.

Since Limesurvey seems trying to attract more business users for the app (see the professional menu in the top right position) I wonder what is the focus on features vs. workarounds. The big commercial systems are getting down in prices (e.g. flat rate for 1500 EUR each month) so the make or buy decisions are getting tougher. The feedback capturing on websites is more and more distributed between pages and few questions like Uservoice or Getsatisfaction.
The positioning of Limesurvey will be cruical. The developers might tell us a bit more about their motivation and aims. Perhaps we have to wait till V2.0 to see ThirdPartyDevs joining the game and show some ways to get sponsors for new applications.
atiut wrote: Still waiting for an example which allows me to see that my feature request for an questiontype "End of Survey" is crap.

Thanks again. Let's collect more insights. Perhaps V2.0 will allow me to contribute more than silly boardposts.

jelo wrote: