Pivot+ Design with condition

[thiemo_goettingen]

Hi Experts:

I am trying homogeneous pivot plus design. I will have around 2500 participants. The code is shown at the bottom. The problem is we can´t find a design considering attributes and conditions. I have two attributes with 5 levels and one attribute with 3 levels plus price (5 levels). We want to restrict one level of one attribute from coinciding with two levels of another attribute (both attributes have 5 levels).


1. It seems pivot design does not work with 'condition' or 'reject'. In our design, we want to set a condition: 'if(alt1.reg>4, alt1.label<>2),if(alt2.reg>4, alt2.label<>2)'. Otherwise there will be unrealistic alternatives. Ngene came to a design but did not consider the conditions.

I am quite new to stated choice experiment and never used pivot before.

I would be very greatfull for any help!
Thank you.

Thiemo

Code: Select all

Design
;alts = alt1*, alt2*, SQ, NC
;rows = 9
;eff=(mnl,d)
;cond:
if(alt1.reg>4, alt1.label<>2),
if(alt2.reg>4, alt2.label<>2)

;model:
U(SQ) = aSQ[-1] + regb[1.7]*reg.ref[1] + prodb[1]*prod.ref[1]+ labelb[1.5]*label.ref[1]+priceb[-0.3]*price.ref[3.99]/
U(alt1) =  regb[2.0]*reg.piv[1,2,3,4,5] + prodb[0.8]*prod.piv[1,2,3]+ labelb[1]*label.piv[1,2,3,4,5]+priceb*price.piv[4.99, 5.99, 6.99, 7.99]/
U(alt2) =   regb[2.0]*reg.piv[1,2,3,4,5] + prodb[0.8]*prod.piv[1,2,3]+ labelb[1]*label.piv[1,2,3,4,5]+priceb*price.piv[4.99, 5.99, 6.99, 7.99]
$


[Michiel Bliemer]

There are quite a lot of issues with your script. Some of your attributes seem categorical and require dummy coding. Your priors are problematic. The pivots are not correctly specified. The number of rows is likely not sufficient.

I recommend starting with the script below, without any pivots. Once you have generated a design, you could translate this back manually into a pivot design if you wish, but note that pivots with categorical variables is generally complicated.

Code: Select all

Design
;alts = alt1*, alt2*, SQ, NC
;rows = 18
;block = 2
;eff=(mnl,d)
;alg = mfederov

;require:
sq.reg=1,
sq.prod=1,
sq.label=1

;reject:
alt1.reg>4 and alt1.label=2,
alt2.reg>4 and alt2.label=2

;model:
U(alt1) = regb.dummy[0.1|0.2|0.3|0.4]   * reg[2,3,4,5,1]                 ? 1=... (base), 2=..., 3=..., 4=..., 5=...
        + prodb.dummy[0.1|0.2]          * prod[2,3,1]                    ? 1=... (base), 2=..., 3=...
        + labelb.dummy[0.1|0.2|0.3|0.4] * label[2,3,4,5,1]               ? 1=... (base), 2=..., 3=..., 4=..., 5=...
        + priceb[-0.1]                  * price[4.99,5.99,6.99,7.99](4-5,4-5,4-5,4-5)
        /
U(alt2) = regb.dummy            * reg
        + prodb.dummy           * prod
        + labelb.dummy          * label
        + priceb                * price
        /
U(SQ)   = aSQ[-0.1]
        + regb.dummy            * reg
        + prodb.dummy           * prod
        + labelb.dummy          * label
        + priceb                * price_sq[3.99]
        /
U(NC)   = aNC[0]
$

Michiel