three questions, many thanks

This forum is for posts covering broader stated choice experimental design issues.

Moderators: Andrew Collins, Michiel Bliemer, johnr

three questions, many thanks

Postby juanprice » Fri Feb 08, 2019 8:24 pm

Dear all
I hope you can help me. I just started working with this amazing software (thanks professors Bliemer, Rose, Collins and team). I have three questions:

1. If I write * to avoid dominant alternatives (alt1* and alt2*) do I also have to use the * in the SQ alternative?
2. Ngene has been running for an hour and the last 40 mins has been stuck in the last iteration. Can I just double click that row and choose that exercise or I need to wait till the software stops completely?
3. If the answer to 2 is that I can just choose that last row (even when the software is still "running") I wonder why in the resulting design my "require" specifications for the "SQ" are not respected by the software (there are some choice tasks that have tax=0 or tax=2 even when I specified a required sq.tax=1). This is my code:

Design
;alts = alt1*, alt2*, sq
;rows = 18
;block = 3
;eff = (mnl,d)
;con
;cond:
if(alt1.tax <> alt2.tax, alt2.fee <> alt1.fee or alt2.fee = alt1.fee) ,
if(sq.tax <> alt2.tax, alt2.fee <> sq.fee or alt2.fee = sq.fee) ,
if(sq.tax <> alt1.tax, alt1.fee <> sq.fee or alt1.fee = sq.fee)

;require:
sq.vis = 1 and sq.colpre = 1 and sq.fee = 1 and sq.tax = 1

;model:

U(alt1) = A1[0.001]*vis[0,1,2] + A2[0.001]*colpre[0,1,2] + A3[-0.001]*fee [0,1,2] + A4[-0.001]*tax[0,1,2]/
U(alt2) = A1*vis + A2*colpre + A3*fee + A4*tax /
U(sq) = A1*vis + A2*colpre + A3*fee + A4*tax + A5[0.01]
$


Many thanks friends
Juan
juanprice
 
Posts: 2
Joined: Fri Feb 08, 2019 5:35 am

Re: three questions, many thanks

Postby Michiel Bliemer » Sat Feb 09, 2019 9:32 am

1. Yes you would likely want to add * to the SQ alternative as well. Note that adding dominance checks as well as other constraints make it more difficult to find feasible solutions that satisfy all your constraints. Looking at your syntax, I think you are overconstraining your design and a feasible solution may not exist that satisfies all your constraints. That means that you may want to change your design dimensions, e.g. increase the number of attribute levels, or alternatively remove constraints that are not strictly necessary.

2. Ngene will never terminate because there are typically billions of possible combinations to check. If the D-error no longer improves at some point, you can terminate the algorithm and take that design.

3. ;cond constraints are compatible with the default swapping algorithm, while ;require constraints are compatible with the modified Federov algorithm. You cannot include both types of constraints. If you use ;alg = mfederov then Ngene will consider the ;require constraints but you will have to remove your ;cond constraints. Note that you can usually rewrite constraints in either the ;require/reject form or the ;cond form.

Michiel

NB: Please post any further questions regarding Ngene in the Ngene forum, this forum is only for general design questions.
Michiel Bliemer
 
Posts: 1705
Joined: Tue Mar 31, 2009 4:13 pm


Return to Choice experiments - general

Who is online

Users browsing this forum: No registered users and 3 guests

cron