netsed logit with panel data

[高佳]

Hi all,

This is the first time that I encountered with choice model in my research, the data we have is discrete choice data at transaction level, further, each customer purchased multiple times, therefore, this is a customer - time panel data.

Besides, the choice structure is well nested, customer first chooses whether to use self-checkout or cashier checkout (only pay with card is available), if self-checkout is chosen, then customer needs to decide whether to pay with smartphone or card. We have no alternative specific variables, only transaction level variables.

Now we want to estimate whether the number of people around at the time of purchase impact focal customer's checkout and payment choice.

My question is, can we run nested logit with customer fixed effect? Or more generally, is there a version of panel data nested logit?

If we can not run a nested logit with customer fixed effect directly, I have two more questions:

1, is there any theoretical background that prohibits people from doing so?
2, is there any other way that we can nested logit to estimate the data we have while kind of take care of customer fixed effect?

your time and knowledge will be deeply appreciated.

JAY

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. We generally see names written with English characters.

You have a sequential decision process. Whether you want to fully model the sequence or not may depend on your research. I'd look hard at nlogit. I don't know if it handles your problem or not. While one might think including dummies for individuals might solve the fe problem, this is not true in conventional logit so is also probably not true in nested versions.
You might also look at https://www.statalist.org/forums/for...ed-logit-model

[高佳]

Hi Phil, thanks for your reply and warm reminder, very helpful. I did not post any code or data is because the question here is a general one: can we run nested logit with individual fixed effect? I read from various resources that this seems impossible for now, however, none of them give the theoretical explanation of why. It would be great if you or any other here can help to clarify this. I think I'm not the first and the last who have this question, clarify it will also help those who run into this later.

Best

JAY