Nested Choice Model

In the logit model, we assumed independent error term with extreme value distributions.

make sense？

Why？

•The i.i.d assumption essentially created the IIA property of Logit model. •This is not completely correct, because other distributions for the unobserved, say with normal errors, we would not get IIA exactly, but something pretty close to it. •We need to relax the independence between the unobserved error term ε

####Solution to IIA

image

Nested Logit Model

image

image

簇与簇之间是独立的，但是簇的内部有相关性

image

image

image

Properties of Nested Logit Model

image

An Example

image

image

image

Decomposition of Nested Logit Model

image

image

image

image

Ink 是内部方案的吸引力

An Example

image

image

Model Estimation

image

image

image

一般用前者，渐进有效；后者一致但不有效

Three-Level Nested Logit

image

image

image