16
Table 3: Simulated ML estimates from choice experiment, mixed logit
|
Bulls traits |
Mean coefficient |
Standard deviation |
|
Random parameters in utility function |
0.4556*** |
0.6684*** |
|
Traction potential |
(0.0842) 1.6838*** |
(0.0812) 1.4233*** |
|
Fertility |
(0.1302) 0.3747*** |
(0.1181) 0.7235*** |
|
Liveweight in Kg |
(0.1209) 0.2772*** |
(0.1053) 0.3650*** |
|
Purchase price (US$) |
(0.0956) -0.2319* |
(0.0529) 0.3087*** |
|
(0.1278) |
(0.2425) |
|
Non-random parameters in utility function Low watering frequency 0.1831*** (0.0501) Dark coat colour -0.0278 | |
|
Constant |
(0.0535) -2.4696*** |
|
(0.2792)_________________________________________ | |
|
Simulated log likelihood at convergence |
-1231.072 N=2783 |
|
Likelihood ratio testa 941.8 (χ02.99(20) = |
37.6); McFadden R2b = 0.312; Halton draws=100 |
Cows traits
Random parameters in utility function
|
Purchase price (US$) |
-0.0333* |
0.1161*** |
|
Trypanotolerance |
(0.0145) 1.5671*** |
(0.0279) 2.1548*** |
|
Supplementary feeds |
(0.0988) -0.4217*** |
(0.1933) 0.7299*** |
|
Milk yield |
(0.0908) 0.4429*** |
(0.1656) 0.4422*** |
|
Reproduction ability |
(0.0741) 0.4866*** |
(0.1294) 0.7166*** |
|
(0.0965) |
(0.1744) |
|
Non-random parameters in utility function | |
|
Low watering frequency |
0.0991 |
|
(0.0606) | |
|
Dark coat colour |
-0.0359 |
|
(0.0505) | |
|
Liveweight in Kg |
0.2911*** |
|
(0.0942) | |
|
Constant |
-2.4601*** |
|
(0.3169)_________________________________________ | |
Simulated log likelihood at convergence -1391.685 N=3036
Likelihood ratio testa 793.6 (χ02.99(20) =37.6); McFadden R2b =0.222; Halton draws=100
a The likelihood ratio test is given by 2(Lβ-Lω), where Lω is the unrestricted maximum log-likelihood from the
mixed logit estimation and Lω is the restricted maximum log-likelihood from the multinomial logit estimation. It
has an asymptotic χ2(k) distribution where k is the number of required restrictions.
b McFadden R2 is computed as R 2=1- Lω Lω
***, **, * indicate that coefficients are statistically significant at the 1, 5 and 10 % levels, respectively, using
P-values in maximum likelihood estimation. Robust standard errors in parenthesis.