Hello, I have a tricky problem to solve. I have a questionnaire on food habits and 3 groups of respondents to randomly allocate to the following groups: vegetarians, beef eaters and chicken eaters. I included a screening question that asks respondents if they eat vegetables, beef or chicken that I need to exclude 1 or 2 groups. For example: 

- if the respondent eats vegetables and chicken only: assign one between vegetable group or chicken group randomly
- if the respondent eats vegetables and beef only: assign one between vegetable group or beef group randomly
- if the respondent eats beef and chicken only: assign one between beef group or chicken group randomly
- if the respondent eats vegetables only: assign vegetable group (no randomization)

Ideally I want these 3 groups to be balanced. Any idea?

Thanks in advance

Hi,
please send a lss export of your survey, respective these relevant questions/groups.
This can be done by a few equations and relevance equations.

But I have not time to create a sample survey.

Ideally I want these 3 groups to be balanced

How do you imagine this?
In the worst case there are no respondents who eat beef. So, what can you do?
Of course you can change the probability to be thrown into vegetable vs. chicken from 50:50 to 80:20, if the chicken group is nearly filled.
But this is something we see later with your survey.

Joffm

By the way: I assume, if he eats all three, he will be assigned randomly to two of the groups.

 

Here's an overly simplified IIS file of the survey. There are 4 groups all with one question: the screening, beef, chicken, and vegetable

With respect to your question, I can't be sure to balance the three groups equally. We will draw our sample from a nation-wide representative panel, so I would anticipate that we have enough respondents to cover all food preferences. My initial idea was to set a maximum N for each group and when the maximum is achieved the group is not shown anymore, only the other 2 remains. Or do it manually by removing one group on the fly. This is another issue to look at when the the random assignment is fixed...

Thanks a lot
Gianluca

Hi, Gianluca,
here your survey back.
You seem to work with an old version of LimeSurvey (2.50 - 2.73).
So I used 2.73.1.

 

 My initial idea was to set a maximum N for each group and when the maximum is achieved the group is not shown anymore

To do this you need to have access of all other responses.what is not possible in older versions of LimeSurvey.
There is a plugin "getStatInSurvey" for 3.x., in 5.x. such a function is built-in.

So, you may adapt the first "join" function (remove the not necessary part.)

Joffm
 

Thanks a lot Joffm,

That was really helpful. Actually I am working with Limesurvey browser, not desktop, so I am not sure what version i have. I'll check out whether I can access the newest version! Thanks

Ggrilli

There is no "desktop" version of LimeSurvey, it is all web-based.

You should see the version details at the bottom right of admin screens.