ghzstudio 发表于 2017-7-7 10:42 
不敢苟同啊~
仔细看一下楼主贴的这两段,其实后面的内容也是不一样的。
前面那段,英文那个论文是两级阶梯 ...
您可以再接下来读一下文章其余部分,我再贴一段,如果要用别人的模型或者拓展,我认为应该是自己理解后,结合自己的文章重新搭建吧?
这样的原文翻译不合适吧
An important aspect of this model is that the household-level demand error,
ε, is heteroscedastic. This occurs because the stochastic term in the household-level demand
specification is the sum of the stochastic terms associated with the K appliance utilization
equations (8). Specifically, from equations (8) to (10), the variance of the household-level
stochastic term is a function of the appliances owned:
var(ε) = XK j=1 Xk K=1 d jdk cov(ε j,εk) (11)
≡ σ(d1, d2,..., dK)2.
We think of the appliance-level stochastic terms as reflecting households’ idiosyncratic tastes
for utilizing appliances. A variety of behavioural considerations then suggest that the covariance
terms entering (11) will tend to be positive, so that the variance of the household-level stochastic
term will increase with the number of appliances owned.
From an econometric perspective, equation (11) is a simple model of group-wise
heteroscedasticity in which the “group” is a specific portfolio of household appliances. Normally,
this would not be a major concern for estimating the parameters of a linear demand specification
such as (10). When consumers face nonlinear prices, however, the variance of the household-level
stochastic term affects the likelihood that a consumer will fall on one tariff segment or another.
This can be seen immediately from equation (6), where (the root of) the variance term, σ , enters
the conditional expectation function and the tariff segment probabilities. The heteroscedastic
variance of unobserved tastes thus affects expected consumption calculations and estimation of
all the demand parameters