choice-metrics.com

Dear Sir,

Albeit, the function of the blocking is to minimize or to show smaller designs to the respondents from factorial design or full choice tasks. Is it possible that, within a Block there exists repeated choice task (levels are appearing more than one within the same block for all attributes) ?
In other words, same rows within a block whether it be a efficient design or orthogonal ?

Repeated choice tasks can occur in both orthogonal and efficient designs if you do not create your design correctly.

When you have unlabelled alternatives (e.g., option A, option B, option C) with generic attributes and levels across all alternatives, you have to generate the design in a specific way.

For orthogonal designs, you need to adopt a sequential orthogonal design strategy. For example, in Ngene you would use ;orth = ood or orth = seq.
For efficient designs, you need to indicate that the alternatives are generic such that repeated choice tasks can be avoided. For example, in Ngene you would specify ;alts = optionA*, optionB*, optionC*, where the * indicates that these are generic alternatives and hence Ngene checks for repeated choice tasks.

If repeated choice tasks do not exist within the design, they also do not exist within a block.

I hope that answers your question?

Michiel

Yes, Thank you sir. But, in doing so the following error persists: "A valid initial random design could not be generated after approximately 10 seconds. In this time, of the 122599 attempts made, there were 0 row repetitions, 6251 alternative repetitions, and 116348 cases of dominance. There are a number of possible causes for this, including the specification of too many constraints, not having enough attributes or attribute levels for the number of rows required, and the use of too many scenario attributes. A design may yet be found, and the search will continue for 10 minutes. Alternatively, you can stop the run and alter the syntax".

Note: I have a qualitative text as level for an attribute such as: Slow, moderate,Fast which i have coded as 1,2,3.

To clarify regarding the term repeated choice task,

I have 3 alternatives and the levels of a certain alternative is repeated ( for example; attribute Levels of Task 4 alt 3 of the 1st block is repeated on the 2nd block of Task2 alt 1)sometimes within the block and sometimes with other blocks.

Sorry it is still not clear, do you mean "profiles" instead of "choice tasks"?

To clarify terminology:

Route A: 5 min, $2
Route B: 10 min, $1

(5 min) and (10 min) are attribute levels.
The combinations of attribute levels (5 min, $2) and (10 min, $1) are profiles.
The combination of profiles (5 min, $2, 10 min, $1) is a choice task.

Repeated attribute levels are common and fine.

Repeated profiles are common and fine, it is fine if profile (10 min, $1) appears within another choice tasks.

Repeated choice tasks are not ok. So you do not want (5 min, $2, 10 min, $1) to be repeated or that (10 min, $1, 5 min, $2) appears as a choice tasks since the profiles are simply swapped and represents the same choice task. Ngene automaticaly avoids such repeated choice tasks, including swapped profiles.

Michiel

Thank you sir, this is much clear now. An error is persisting in Ngene as it says :

"A valid initial random design could not be generated after approximately 10 seconds. In this time, of the 122599 attempts made, there were 0 row repetitions, 6251 alternative repetitions, and 116348 cases of dominance. There are a number of possible causes for this, including the specification of too many constraints, not having enough attributes or attribute levels for the number of rows required, and the use of too many scenario attributes. A design may yet be found, and the search will continue for 10 minutes. Alternatively, you can stop the run and alter the syntax".

I have a qualitative text as level for an attribute such as: Slow, moderate,Fast which i have coded as 1,2,3. Is it because they have been coded 1,2,3 ?

It means that the vast majority of the generated designs contain a dominant alternative (116,348 out of 122,599 evaluations). Your design is therefore heavily constrained and the default algorithm in Ngene, which maintains attribute level balance, cannot find a design without any dominant alternatives.
You will need to switch to the modified Federov algorithm by adding:

;alg = mfederov

Note that this algorithm does not maintain attribute level balance and therefore you may wish to include additional attribute level balance constraints.I can assist if you are happy to share your complete Ngene syntax here.

Michiel

Dear Sir,

I have managed to do in Ngene and it gives me these results :

D error 3.5E-05
A error 0.000298
B estimate 100
S estimate 0


Note: I have coded the level of an attribute as 1(slow),2(moderate),3(fast)- since the attribute is a qualitative text.

Here is the Ngene code,

design
;alts = Tariff1*, Tariff2*, Tariff3*
;rows = 16
;eff= (mnl,d)
;block= 2
;model:
U(Tariff1) = bchargingspeed * Chargingspeed[11,50,150]
+ bdist * Distance[500,1000,1500]
+ bcost * Price[35,55,75]
+ brwt * RWT[0,10,20]

/
U(Tariff2) = bchargingspeed*Chargingspeed+bdist*Distance +bcost*Price +brwt*RWT /
U(Tariff3) = bchargingspeed*Chargingspeed+bdist*Distance +bcost*Price +brwt*RWT
$

Am i doing anything wrong? Your advise is highly appreciated.

You can add priors to indicate the preference order of the attribute levels to avoid dominant alternatives, see code below where I guessed some of the near-zero priors to indicate preference orders.
You can use dummy coding for categorical variables, see below.
If you use the modified Federov algorithm, you can add attribute level constraints such as requiring that each level appears between 4 and 6 times across the design, see below.

Code: Select all

design
;alts = Tariff1*, Tariff2*, Tariff3*
;rows = 16
;eff= (mnl,d)
;alg = mfederov
;block= 2
;model:
U(Tariff1) = bchargingspeed.dummy[0.0001|0.0002] * Chargingspeed[2,3,1] ? 1 = slow (base), 2 = medium, 3 = fast
           + bdist[-0.000001]                    * Distance[500,1000,1500](4-6,4-6,4-6)
           + bcost[-0.000001]                    * Price[35,55,75](4-6,4-6,4-6)
           + brwt                                * RWT[0,10,20](4-6,4-6,4-6)
           /
U(Tariff2) = bchargingspeed*Chargingspeed+bdist*Distance +bcost*Price +brwt*RWT /
U(Tariff3) = bchargingspeed*Chargingspeed+bdist*Distance +bcost*Price +brwt*RWT
$

Michiel

Thank you professor.

In addition to this, i have a question regarding the manual editing of Experimental design (in case if a surveyor observe dominant profile in a choice task).

For example,

How important it is, for the repetition of the attribute levels in an experimental design?
The appearance of the attribute level should be eventually distributed (attribute level appear equally often throughout the choice tasks) in case of the efficient design?

Does it impact estimation of the main and interaction effects in my logit model?