Truly randomization with even distribution splits

More
1 month 2 weeks ago #160186 by addisonresearch
addisonresearch created the topic: Truly randomization with even distribution splits
Hello,

We have a requirement where we have 4 categories and each category contains 2 questions separated by slash(/) and we have to split in such a way that half of the respondents are asked the first half of the question and the other half are asked the second part of the question. For example:

Investing in transit services, including an expansion of Toronto subway lines to Richmond Hill and additional parking at Go stations / Investing in skills, training, and re-training programs

50% of the respondents should be asked: Investing in transit services, including an expansion of Toronto subway lines to Richmond Hill and additional parking at Go stations

And the other 50% should be asked: Investing in skills, training, and re-training programs

Technically there are 8 questions, but each respondent will only be asked 4 of those 8.


I’ve include the full list of questions below.


(SPLIT) Cutting taxes on the middle class and raising them on the top 1% / Strengthening the Canada

Pension Plan

(SPLIT) Investing in Canada’s military and the men and women of the Canadian Armed Forces /

Introducing the Caregivers tax credit, to help Canadians care for sick and elderly family members

(SPLIT) Investing in affordable housing / Legalizing marijuana

(SPLIT) Investing in transit services, including an expansion of Toronto subway lines to Richmond Hill

and additional parking at Go stations / Investing in skills, training, and re-training programs

Please let me know if there is any way to make it feasible?

Thanks,

Please Log in to join the conversation.

More
1 month 2 weeks ago #160191 by Joffm
Joffm replied the topic: Truly randomization with even distribution splits
Well, the easy answer:
create four random numbers (1,2) R1-R4 and select each of the four splits according to this number.
If you see during the survey that the distribution isn't quite equal just change the relevance of the questions.
Therefore I like to use a random number (1-100) and can fine tune by saying something like "Split1,if random<54", "Split2,if random>53".

Because I do not know what your "Truly" means the more sophisticated way, if you have a participant table:
There are only 16 different splits.
Before starting the survey add an additional attribute to each respondent (1-16), randomly by dices, by EXCEL, by your grandmother or whatever.
Add the relevance to your eight questions.
This way you can be sure that you have an equal distribution - only if everybody completes the survey.

Best regards
Joffm

Please Log in to join the conversation.

Start now!

Just create your account and start using Limesurvey today.

Register now
Join our Newsletter!