Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

Block and Quota

Hello, I know similar questions have been asked before, but I am having trouble combining quotas requirments with three different blocks of questions.

I have three blocks of information/DCE questions e.g.: Block1 : Q1-Q10, Block 2: Q11-Q20, Block 3: Q21-30. I want respondents to only see one block. And everyone meets back at Q31.

But within each block, I want quota limits on age and gender. But these quotas are the same for each block; all blocks will have 50:50 male/female split and 25/25/25/25 for ages 18-25/26-35/36-55/56+

My question is: How do I ensure that respondents are first screened based on age and gender and then put into the block that still needs the quota target?

And then is there a way to call which block they were assigned in the skip logit section? Such that if you saw Block 1 you might need to skip Q32.

Thank you for your help!
asked 6 days ago by Keila
Maybe the methodology guys on this forum will have a nicer solution, but would one option be to create a quote question with 24 cells?  So the first one would have logic like "Sex = 1 And Age = 1," then the second "Sex = 1 And Age = 2," and so on for the first eight.  Then those eight could be repeated two more times to represent the three blocks.  Finally, the quota question could be set to "Least Fill" and respondents could be skipped to one of the three blocks depending on which quota cell they fit into.

As for the second part of your question, you can always use the name of a quota question in skip logic similar to other question / variable names.  This will return the cell number that the respondent landed in (or a special code if they did not make it into any cell); you could use this to determine which block they were sent to.  Or, if you find it easier, you could use the Sawtooth Script function Shown to determine whether they saw Q1, Q11, or Q21.  Or, as a final option, you could add a pass-in field to your survey and set it to 1, 2, or 3 after assigning respondents to their block.

1 Answer

0 votes
Hi Keila,

You can first have age and gender question. Based on the response in these questions, you can have AgeAndGenderQuota that will have 2 (Male/Female) X 4 (4 age groups) X 3 (since you have 3 blocks) = 24 cells.

24 cells can be as below:

First 8 cells for Block 1

Block1MaleAge1
Block1MaleAge2
Block1MaleAge3
Block1MaleAge4
Block1FemaleAge1
Block1FemaleAge2
Block1FemaleAge3
Block1FemaleAge4

Similarly 8 + 8 cells for other two blocks. So, each cell will be given the limit that you want to set (in proportionate). Change the membership method to Least fill (percentage). Each respondent will qualify for 3 cells out of 24 cells but because of least fill, respondent will qualify for only one cell.

Now for asking the blocks, you can set a constructed list whose parent list will have 3 members (Block1,Block2,Block3)with the below logic:
1. If quota value = 1 to 8, then Block 1
2. If quota value = 9 to 16, then Block 2
3. If quota value = 17 to 24, then Block 3

Now click on the Randomized button and select Blocks. Select Add Set and then Add Block. Define 3 blocks with start and end anchors.
Select the Use Constructed List button and choose the constructed list you have created.

Similarly, you can set skip on Q32 based on value in quota question.

I hope this helps.

Regards,
Rahul Mittal
answered 6 days ago by Rahul Mittal Silver (8,860 points)
Hi Rahul, Thanks for your reply. I see what you have done here. The problem is that age and gender were only two examples. I actually have 5 quota variables each with quite a few categories. So with three blocks this would be  a huge number of cells (11,000+)

I was wondering if I could: set the female quota for the entire sample (i.e. if I want 600 and 200 in each of the 3 blocks)
Then assuming they meet the socio-economic quotas, they randomly get assigned to one of the three blocks, using the method from the link below.

https://legacy.sawtoothsoftware.com/forum/14205/randomly-show-ad-to-be-rated-the-quota-for-that-ad-is-not-filled?show=14205#q14205

I guess I run the risk of not having each block the exact required quota, but if they are randomly being allocated then it should be close.... would that work? Otherwise I would need to simplify your steps above such that there were no interaction terms: i.e. gender (2) * 3 = 6 cells; age (4) * 3 = 12 cells

But some categories like education and income have 8 categories each.. this is just so many quota cells....
I have understood your approach of assigning a respondent a block. But using this approach, it is not obvious that you will have equal counts across each combination (if you really want to have approximate equal counts in each possibility) and it is also not sure that the counts will be near to each other - it totally depends in what order you are getting the sample of each combination. If you are OK with having counts not in the required proportion, your approach is good to go.

I also understand that 11000+ is a great number and have no idea if the sawtooth will be able to adapt this many number of cells (means it will not hang) easily in a single quota. You can check this with sawtooth support team if having 11000+ cells in a quota will work smoothly or not, and yes if they approve you on that, please use the import csv functionality of quota to apply conditions.
...