Individual part-worth estimates

[JoshSchramm94]

Dear John,

first of all, thanks a lot for this amazing package. I was wondering if there is the chance to extract individual part worths similar to the indpar from the mlogit package?

Thanks a lot

[jhelvy]

Thanks! You're certainly not the first to ask for this, so I'll probably need to include it at some point. This issue also mentions individual parameters (but in the context of prediction) and it includes most of the math needed to implement it. It's really just a matter of sitting down and writing the code - it's all been well-established with mixed logit for a long time. I just have other projects that are consuming more time. I'm also hoping someone else will come along and give it a shot themselves and post their code to make adding this feature a lot less work for me :). So, it will be done at some point, but I'm not sure when that will be. If I can get to it by the end of August then good, otherwise it'll probably wait a while once I start teaching again in the fall.

[JoshSchramm94]

I see, thanks a lot for your detailed response. :-)

[likever20212]

Hi, Johh. I'm so pleased that the package is faster than other packages running mixed logit model. However, the individual coefficient of each random variables could not available like indpar in mlogit package yet. Have I miss this function or the function is developing? Thank you so much!

[jhelvy]

I have not yet implemented it. Work on logitr has paused for a while for me while I develop other packages (most recently surveydown, which is quite a big project now). If anyone wants to work out the code to implement this I would be happy to review a PR. In the meantime, it'll have to wait until I have to time to work on it. I have plans of possibly taking my sabbatical in a year or two from now to focus on software development, where I could maybe implement this at that time.