Conditional logit (clogit) in Julia?

Specific Domains Statistics
1

Given it isn’t at all my subject, a colleague of mine interested in Julia asked if there is a Julia package that implements the “conditional logit” model, similar to the clogit function in R survival package.

Having browsed the subject a bit, it seems that the conditional logit is “equivalent” (??) to a “Cox model with a particular structure” and that Cox models are implemented in the fit(::Type{CoxModel},…) function of the Survival.jl package.

Is that right? Can the fit function of that package be used for a conditional logit model ? Is it straight-forward ? Does anyone already did and can share some simple code I can show to my colleague ?
Thanks !

2

he is still asking :wink:
A new PhD student would like to try using Julia but my colleague (that will be his supervisor) says that there isn’t nothing about this in Julia… is this still true?
They will work on choice modelling, perhaps GitHub - mattwigway/DiscreteChoiceModels.jl: Discrete choice/random utility models in Julia will be relevant ?

3

There may be specific code available from someone, though I am not aware of any. This model is easy to program, I did it long ago using the Gauss language :roll_eyes: I have some generic MLE code at Econometrics/src/ML at main · mcreel/Econometrics · GitHub. If one were to write the conditional logit likelihood function using the form in the examples in the Likelihoods subdirectory, that would work.

2 Likes
4

Maybe @nosferican?

5

I suspect having seen it in the ecosystem. A quick search led me to GitHub - alancrawford/GEV.jl: Estimate Generalized Extreme Value Models It may be worth just checking a couple examples from the R package and seeing if the functionality is suitable for their purpose.

2 Likes