Zufallsauswahl with equal distribution?

[Alan Galarza Velarde]

Hello everyone,
I would like to use the random selection (Zufallsauswahl) in order to ensure that each participant only sees 1 of 4 pages. Will this be equally distributet (e.g. 1000 participants = 250 participants per page)? Ty for your help

[Michael T]

I am not part of unipark.

 

According to the pdf-manual from 2016 (yeah, quite a bit outdated but cant find anything about this in the online manual) its NOT. Page 280, headline 6.2.2 clearly says “Bitte beachten Sie: Es handelt sich um eine Zufallsauswahl, nicht um eine Gleichverteilung.” – roughly translated as “Please note: This is a random selection, not an uniform distribution”

 

I would suggest using a user defined variable filled by a random trigger (execution position: direct, execute trigger in preview: yes, execute trigger multiple times: no, values from 1 to 4). Then you have to create four filters using the user defined variable and adding a page to each filter. A random trigger has to option to aim for uniform distribution. But it is unlikely that you’ll get exactly 250 participants for each bucket. You could end up with something like 248/249/253/250.

 

Why? Lets assume we have just two buckets and want three participants per bucket. From what I know the number is added to the bucket when the participant finishes the questionnaire. Two people start the questionnaire and both get the number “1” from the random trigger (random number generator or rng). Another two people start the questionnaire and get “2” from rng. None of those four has finished the questionnaire and now three participants entering the questionnaire. So the rng could assign them 1/1/2 and if all 7 participants finish the questionnaire, you’ll end up with 4/3. But it could also happend that one of those with the number “2” doesnt finish the questionnaire and you’ll have 4/2 at the end.

 

Also have a look at

https://qbdocs.atlassian.net/wiki/spaces/DOC/pages/1441563760/Random+Trigger

checking the last paragraph “Uniform distribution”.

 

Would be great for you (and me) if Unipark could confirm that.

[Carlotta Mayer]

Hey Micheal T,

ty for the quick reply I did try it with the trigger you mentioned and it seems to work so ty.

 

But you are right it would be nice to get it confirmed from unipark and I am wondering why its not equally distributing.

[Birgit Putz]

Hello Carlotta, hello Michael T,

 

A random selection only provides an approximately equal statistical distribution with very large numbers of participants. Very large means more than about 10,000 participants for each characteristic.

 

If you want to achieve true equal distribution, you have to use quotas.

 

Best regards,

Birgit

[Birgit Putz]

Hey Birgit,

 

ty for the reply I don’t really understand why you suggest quotas instead of a trigger. When I understood the quotas correctly I would have to define from the start when which participant is going through a specific filter? So this would not be random but dependend on a specific variable (question?).

 

The trigger aimes for an equal distribution and only slightly varies if I understand it correctly so this way I could achieve a random equal distribution isn’t that correct?

 

Best regards
UpvoteReply

[Birgit Putz]

Dear Carlotta,

 

Quotas can also be used to achieve equal distribution. For this purpose, the quotas are defined somewhat differently than, for example, if you want to count men and women or age groups.

 

With an equal distribution via quotas, the participants are assigned to the groups one after the other, and then again from the beginning, e.g.

Group 1

Group 2

Group 3

Group 1

Group 2

…

 

In this way, an almost (!) one hundred percent equal distribution is achieved even with a small number of participants.

 

If you are interested, you can find some examples here in the forum. Mostly in German.

 

Otherwise feel free to ask again, but better with a new request.

 

Best regards,

Birgit

[Author]

Posts