英语翻译Alternative specifications of qualitative choice modelsa

英语翻译
Alternative specifications of qualitative choice models
are based on the probit and the logit models (Greene,
1998).These models are used to denote a regression model
in which the dependent variable is a binary variable taking
the value of 1 if the event occurs and 0 otherwise.The
distinctiveness of the probit and logit models is that the
predicted value of the dependent variable is interpreted as
the probability that the event happens.The difference
between both models can be found in the distribution
function associated with each model.
In principle,because logit uses the binomial distribution
and probit uses the cumulative normal distribution,one
should use logit if one assumes the categorical dependent
variable reflects an underlying qualitative variable and use
probit if one assumes the dependent reflects an underlying
quantitative variable.However,in practice,these alternative
assumptions rarely make a difference in the
conclusions,which will be the same for both logit and
probit under most circumstances (Ham,Brown,& Jang,
2004).
In any case,when the dependent variable can take more
than one value (not only 0 and 1),but it remains discrete
and bounded,the generalization of the simple logit and
probit models is required.In this way,Greene (1998)
proposes two alternatives.The first one is the so-called
multinomial model which main distinctiveness is that the
dependent variable is a discrete and bounded variable
taking values of 1,2,3,4,5,etc.However,although
multinomial models are applicable to categorical data,
they cannot capture the order that the dependent variable’s
alternative choice follows.For this reason in this study
the second alternative that proposes ordered logit and
probit models has been chosen because,as well as
taking the bounded,discrete nature of the dependent
variable into account,they also take its order into
consideration.
With these models,the implicit initial supposition is that
the latent variable of opinion ðOiÞ is a linear function of
explanatory variables xi and di:
where 2 is a random error and bx and bd are vectors of
parameters to be estimated.
The observed category Li is based on Oi,in accordance
with the following rule: