Labeled design

[PBallco]

Hello,
I have never used a labeled design before and would like some guidance.

These are my variables:
*1. Composition = 1 (wood), 2(hay), 3(rice) (actual situation)
*2. Durability = 1, 2, 3 (actual n/a)
*3. Water = 1%, 2%, 3% (actual 0%)
*4. Price = 11, 22, 33 (actual payment 0)

In this case, it does not make sense to use an unlabeled design as the actual situation in "Composition" cannot be matched with the other attributes and their levels. Therefore, my idea is to create a labeled design where the attributes and levels are compared with the actual situation instead of the no-buy option. It would have been easier if I had one attribute with different levels (e.g., price between the reference and the rest) to generate the labeled design, but unfortunately, the only difference in my case is the actual situation with the rest. For that, I tried to design the following (not surprisingly) without any result:

Design
;alts = com, nobuy
;rows = 18
;block = minsum
;orth = ood
;model:
U(com) = b1[-1] *PRIcom[11, 22, 33]
+ b2.effects[0.9|0.3] *com[1, 2, 3]
+ b3.effects[0.5|0.1] *dura[1, 2, 3]
+ b4.effects[0.8|0.3] *wat[1, 2, 3]
/
U(nobuy) = b0[-2]
$

Q1: Is there another way that I can generate a labeled design with the following information?
Q2: Wouldn't it be easier and correct to generate an unlabeled design, excluding the reference levels mentioned in my variable list from the design, and use the actual situation in each choice task instead of the no-buy option?

Thanks,

[Michiel Bliemer]

I do not quite understand what you mean with "the actual situation in "Composition" cannot be matched with the other attributes and their levels". I also do not understand what you mean with 'actual situation', do you mean a status quo alternative? If so, why do you include an opt-out instead of status quo? Without understanding what you want to do I cannot give any advice.

The Ngene script will not work because ;orth = ood is only for unlabelled experiments, in your case you could use ;orth = sim.

Michiel

[PBallco]

My apologies for the confusion!
The design was successfully executed using 'sim'.

Correct, the 'actual situation' represents the status quo alternative. I am including the opt-out option instead of the status quo because the values of the status quo are all zeros.

In my study, we are examining the type of mulch used at the base of trees compared to trees without any mulch. In this case, the status quo is a tree without mulch, which means no durability benefits, no water conservation, and a zero price.

Q1: Is the design I am using correct?
Q2: Would it be correct to use an unlabelled design (;orth=ood) with the same concept of the status quo as in the labelled design?

Thanks,

[Michiel Bliemer]

You can use unlabelled alternatives for two composting options and then add a labelled status quo alternative, see script below.
For model estimation it is usually helpful if the status quo level also appears as a level in the other alternatives, perhaps for some but not all attributes. But if this is not possible then these will all end up in the constant, which makes interpretation a bit more difficult.

Your prior for b1 was too large for the attribute levels of price. Preferably use priors from a pilot study so that you are sure that they are ok.
You did not indicate the number of blocks, I set it to 3.

Code: Select all

Design
;alts = com1*, com2*, sq
;rows = 18
;block = 3, minsum
;eff = (mnl,d)
;model:
U(com1) = b1[-0.1]            * price[11, 22, 33]
        + b2.effects[0.9|0.3] * com[1, 2, 3]
        + b3.effects[0.5|0.1] * dura[1, 2, 3]
        + b4.effects[0.8|0.3] * wat[1, 2, 3]
        /
U(com2) = b1                  * price
        + b2.effects          * com
        + b3.effects          * dura
        + b4.effects          * wat
        /
U(sq)   = b0[-3]                            ? this constant captures all other attributes
        + b1                  * price_sq[0]
$

Michiel

[PBallco]

Thank you so much Michiel! You are a savior!