some questions about effect coding and WTP, and the minimum

[xiaojin]

Dear Ngene experts:
I am relatively new to Ngene. In the design of DCE, I encountered some questions. I want to find the answers in the forum. If I can get your help, I would appreciate it very much.
This is my syntax:
Design
;alts=alt1*,alt2*,alt3
;rows=6
;eff=(mnl,d)
;model:
U(alt1)=b1.effects[0.0000001]*A[0,1]+b2.effects[-0.0000001|-0.0000002]*B[0,1,2]+ b3.effects[0.0000001|0.0000002]*C[0,1,2]+b4.effects[0.0000001|0.0000002]*D[0,1,2]+ b5.effects[0.0000001|0.0000002]*E[0,1,2]/
U(alt2)=b1*A+b2*B+b3*C+b4*D+b5*E/
U(alt3)=b0[0]
$

And my questions is :
1 As shown above, I use effect coding, but after collecting the data, I will calculate WTP. And my question is whether the choice of dummy coding or effect coding will have an impact on the final calculation of WTP? And what should I pay attention to when using effect coding to calculate WTP?
2 I have doubts about the minimum number of rows. I know that it should satisfy the attribute level balance. So in my syntax, is “rows=6” feasible?
3 And “U(alt3) =b0[0]”, Is it feasible to set the b0 to 0?

If my question seems very ignorant, please forgive me. Thank you again for your help.

[Michiel Bliemer]

1. If you are using dummy or effects coding then your WTP calculations are different than in the case of linear coding where it is simply a ratio of two coefficients. WTP calculations using dummy coding or effects coding are essentially the same, so you can choose either. You should preferably not use dummy or effects coding for continuous variables such as price, since it makes it easier to compute WTP values and it allows for use in forecasting (e.g. if you use $1, $3, and $5 as levels for effects coding, then you cannot apply the model for a price of $4). If you like nonlinear effects in continuous variables, you can consider a continuous transformation, e.g. a logarithm of price. Such transformations also make the WTP calculations somewhat more complex.

2. Yes 6 will ensure attribute level balance, but the number is very small and therefore there is not much variation in your data such that your parameters may be difficult to estimate. I would suggest using at least 12 (and block your design, ;block = 2, i.e. making two versions of the survey).

3. Yes that is fine.

Michiel

[xiaojin]

Dear Ngene experts, Thanks so much for your detailed explanation for how to calculate WTP. I am really grate for your always and great help. The variable of price in my study was set as continous variable. As you explained, WTP is calculated as a ratio of two coefficients. The price variable was continuous. But the other variable was effect coded. Then how to calculate WTP for the effect coded variable? I am really puzzled by that question.

[Michiel Bliemer]

WTP is the trade-off between cost and another attribute. Assuming that cost/price is continuous and the other attribute is dummy or effects coding with L levels, then there will be L-1 WTP values, namely a WTP going from level 0 to level 1, another WTP for going from level 1 to level 2, etc. You will need to compute the difference in utility between two dummy/effects coded levels and based on this difference you compute the WTP.

For example:

Code: Select all

U(alt1) = comfort.dummy[0.1|0.3] * COMFORT[1,2,0]  ? 0 = low, 1 = medium, 2 = high
        + price[-0.2]            * PRICE[2,4,6]

The increase in utility from low comfort to medium comfort is 0.1.
A price increase of $1 decreases utility by 0.2, therefore a price increase of $0.50 decreases utility by 0.1.
In other words, the WTP for medium comfort (relative to low comfort) is $0.50.

The increase in utility from medium comfort to high comfort is 0.2.
A price increase of $1 decreases utility by 0.2.
In other words, the WTP for high comfort (relative to medium comfort) is $1, and the WTP for high comfort (relative to low comfort) is $1.50.

It is not easy to explain here and it is also not an Ngene related question, so if it is not clear then hopefully you can find another source to look this up, possibly in the book of Hensher, Rose and Greene (2015) Applied Choice Analysis.

Michiel