mlogit with fixed effect

[Truong Quoc Phan]

Hello,

I have a question about mlogit/logit as follow:

Model:

mlogit(y) = a + bx + fixed-color

where:
- fixed-color: is a categorical variables that takes three values: blue (1), green (2), and red (3)

Normally, we can run this as:

mlogit y x i.color

What Stata run will assign 2 dummy variables for green and red. Therefore, the result:
- Intercept (which including blue)
- b
- coefficient for green (2) and red(3)

However, I would like a method such that it can extract the coefficient for the fixed effect of blue (1) separated from Intercept rather than included with it. (I want the coefficient for all three colors, not changing the base from blue to other color)

Is there any way/trick to do so?

Thank you

[Erik Ruzek]

I think your best bet is to use margins to get predicted probabilities for each color. If you are not familiar with margins, I'd highly recommend reading Richard Williams' work on the topic. Start with his excellent Stata Journal article. This will help you understand what margins are and the appropriate syntax to get the quantity you want to estimate.

[Truong Quoc Phan]

I think your best bet is to use margins to get predicted probabilities for each color. If you are not familiar with margins, I'd highly recommend reading Richard Williams' work on the topic. Start with his excellent Stata Journal article. This will help you understand what margins are and the appropriate syntax to get the quantity you want to estimate.

Thank you for your reply,
But I tried to do other calculations not related to margin
For those calculation, it requires separated coefficients for three categories of fixed-color

[Richard Williams]

See

https://www.stata.com/links/stata-ba...s%20variables.

In particular, does ib(none) combined with no constant do what you want (see around the middle of the handout).