choice-metrics.com

A little background on the design is that we are looking to calculate the
willingness to pay of a few different products among various attributes.
There is different complete design for each product. I am using the WTP
metric as the efficiency major. For one of the designs everything looks
good and normal, but for the others the estimated willingness to pay that
ngene is reporting is ridiculously inflated.

This first design looks reasonable

Code: Select all

Design
;alts = alt1, alt2, alt3, alt4
;rows = 18
;block = 2
;eff = (mnl, wtp(wtp1))
;wtp = wtp1(*/price)
;con
;cond:
if (alt1.f=[0,1], alt1.a=0),
if (alt2.f=[0,1], alt2.a=0),
if (alt3.f=[0,1], alt3.a=0)
;model:
U(alt1) = label[0.757]*A[0,1] + seller[0.868]*B[0,1] + enviro[1.4]*C[0,1]
+ econ[.477]*D[0,1] + price[-.808]*E[4.5,5,5.5,6.25,7.50] +
location.effects[.642|.455|.346|.0|.0]*F[0,1,2,3,4,5]/
U(alt2) = label[0.757]*A[0,1] + seller[0.868]*B[0,1] + enviro[1.4]*C[0,1]
+ econ[.477]*D[0,1] + price[-.808]*E[4.5,5,5.5,6.25,7.50] +
location.effects[.642|.455|.346|.0|.0]*F[0,1,2,3,4,5]/
U(alt3) = label[0.757]*A[0,1] + seller[0.868]*B[0,1] + enviro[1.4]*C[0,1]
+ econ[.477]*D[0,1] + price[-.808]*E[4.5,5,5.5,6.25,7.50] +
location.effects[.642|.455|.346|.0|.0]*F[0,1,2,3,4,5]/
U(alt4) = b0[-5]$

but this second one is giving me WTP that far outstrips the price of the good

Code: Select all

Design
;alts = alt1, alt2, alt3, alt4
;rows = 18
;block = 2
;eff = (mnl, wtp(wtp1))
;wtp = wtp1(*/price)
;con
;cond:
if (alt1.f=[0,1], alt1.a=0),
if (alt2.f=[0,1], alt2.a=0),
if (alt3.f=[0,1], alt3.a=0)
;model:
U(alt1) = ctgrown[0.515]*A[0,1] + seller[.834]*B[0,1] +
enviro[.272]*C[0,1] + econ[.368]*D[0,1] +
price[-.016]*E[180,200,220,250,300] +
b6.effects[.4|.323|0|.34|0]*F[0,1,2,3,4,5]/
U(alt2) = ctgrown[0.515]*A[0,1] + seller[.834]*B[0,1] +
enviro[.272]*C[0,1] + econ[.368]*D[0,1] +
price[-.016]*E[180,200,220,250,300] +
b6.effects[.4|.323|0|.34|0]*F[0,1,2,3,4,5]/
U(alt3) = ctgrown[0.515]*A[0,1] + seller[.834]*B[0,1] +
enviro[.272]*C[0,1] + econ[.368]*D[0,1] +
price[-.016]*E[180,200,220,250,300] +
b6.effects[.4|.323|0|.34|0]*F[0,1,2,3,4,5]/
U(alt4) = b0[-4.44]$

I would like to point out that should I calculate the WTP by hand with the
priors given for say the CTgrown attribute I have 0.515/-0.016 = 32.1875
while ngene is giving me the number 1486.59. The probabilities and
utilities section does not seem to my untrained eye to be showing me any
problems. Any guidance here would be much appreciated.

Dear nrwright

Ngene is not estimating the WTP values. What it is reporting is the C-error, based on the prior parameter estimates you have assumed. The prior parameter estimates that you input provide your best guess as to wha the true WTP values are in the population, which is different to the C-error which is a summary measure of the sum of the variances of the WTP estimates. Wht the large vaules are telling you is that you can expect huge variances around your WTP estimates, assuming your priors are correct, and hence will require a very large sample size to detect whether in this sample, the WTP values are statistically different from zero or not.

John