Nesting

[Yuri Groeneveld]

Dear Michel,

I am estimating a nested destination-mode choice model but the mu is lower than 1, 0.46. I've read on this google group that thats no correct and the nesting is done wrong. Yet I can't find the error. Thus, is it incorrect or is it possible in some cases?

for i in PC4_limburg:
    nests.append((MU_zone, [int(f'{i}1'), int(f'{i}2'), int(f'{i}3')]))

Here I create my nests. For every zone, I nest the car, pt and bike with the same mu for every zone. Because the nesting is the same for every zone.

Thanks in advance,

Yuri

[Bierlaire Michel]

Indeed, MU must be greater or equal to one. 

It is actually a good iea to constrain the lower bound of the MU to 1. 

I suggest to use a different mu for each nest.

Also, try different nesting structures. Instead of nesting per destination, try to nest per mode. 

If nothing works, consider cross-nested logit. 

--
You received this message because you are subscribed to the Google Groups "Biogeme" group.
To unsubscribe from this group and stop receiving emails from it, send an email to biogeme+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/biogeme/3391781d-d8aa-4811-84df-21e40837929bn%40googlegroups.com.

[Yuri Groeneveld]

Another question,

I can't find clearly how to calculate the probability for each nest and within the nest, only the derivative.

Could you elaborate?

Thanks,

Yuri

Op dinsdag 17 januari 2023 om 15:53:31 UTC+1 schreef michel.b...@epfl.ch:

[Yuri Groeneveld]

Thanks.

When I try nesting with the different modes, car and pt are 1 when constrained to 1 and bike is 1.07.

When I try a CNL the run time takes more than 72 hours. The run time of a mode NL is already over 5 hours.

My alternatives are zones in a province in The Netherlands and I want to apply that mu to other zones in The Netherlands thus I can't have specific mu's for specific zones.

With kind regards,

Yuri

Op dinsdag 17 januari 2023 om 15:53:31 UTC+1 schreef michel.b...@epfl.ch:

Indeed, MU must be greater or equal to one.