how to interpret significance of exp(beta) in size function of destination choice model

[Mariska]

Dear Michel Bierlaire,

I am estimating a destination choice model with size function (in order to include destination attractiveness based on # inhabitants, # jobs and # student spots). Hence, my utility function has the following form (based on the NRM documentation, see attached): 

U = b1x1 + b2x2 + log(b3x3 + exp(b4)*x4)

Now, how should I interpret the significance of b4? Should I still consider the difference from zero (e.g. at 5% significance level (P<0.05)) as provided in biogeme results or does that not hold any longer, since it is added as an exp within a log?

I have added the biogeme part of my code for clearity on the utility function.

I hope you can shed some light on this matter. Thank you very much!

Greetings, Mariska

[Bierlaire Michel]

On 15 Oct 2021, at 17:25, Mariska <m.van.e...@gmail.com> wrote:

Dear Michel Bierlaire,

I am estimating a destination choice model with size function (in order to include destination attractiveness based on # inhabitants, # jobs and # student spots). Hence, my utility function has the following form (based on the NRM documentation, see attached): 

U = b1x1 + b2x2 + log(b3x3 + exp(b4)*x4)

Now, how should I interpret the significance of b4? Should I still consider the difference from zero (e.g. at 5% significance level (P<0.05)) as provided in biogeme results or does that not hold any longer, since it is added as an exp within a log?

By significance, you mean the t-test, I guess. The reported t-test tests the hypothesis that the parameter is equal to zero, which is of no interest to you in this case.

I have added the biogeme part of my code for clearity on the utility function.

I hope you can shed some light on this matter. Thank you very much!

Greetings, Mariska

--
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/6eee3a9e-7c2b-4dca-9933-5ee0889b4fafn%40googlegroups.com.
<Biogeme part of code.py><size function.png>

[Mariska]

Dear Michel,

Thank you for your reply. Indeed, I meant the t-test (I am sorry for not being clear). 

So, I understand now that the t-test with hypothesis that the parameter is equal to zero, is not of interest in this case. However, what hypothesis should I test instead?

Kind regards,

Mariska

[Bierlaire Michel]

On 20 Oct 2021, at 10:43, Mariska <m.van.e...@gmail.com> wrote:

Dear Michel,

Thank you for your reply. Indeed, I meant the t-test (I am sorry for not being clear). 

So, I understand now that the t-test with hypothesis that the parameter is equal to zero, is not of interest in this case.

Exactly.

However, what hypothesis should I test instead?

Well, this is your model. You should know the modeling hypotheses that you want to challenge. 

Kind regards,

Mariska

On Tuesday, 19 October 2021 at 17:15:37 UTC+2 michel.b...@epfl.ch wrote:

On 15 Oct 2021, at 17:25, Mariska <m.van.e...@gmail.com> wrote:

Dear Michel Bierlaire,

I am estimating a destination choice model with size function (in order to include destination attractiveness based on # inhabitants, # jobs and # student spots). Hence, my utility function has the following form (based on the NRM documentation, see attached): 

U = b1x1 + b2x2 + log(b3x3 + exp(b4)*x4)

Now, how should I interpret the significance of b4? Should I still consider the difference from zero (e.g. at 5% significance level (P<0.05)) as provided in biogeme results or does that not hold any longer, since it is added as an exp within a log?

By significance, you mean the t-test, I guess. The reported t-test tests the hypothesis that the parameter is equal to zero, which is of no interest to you in this case.

I have added the biogeme part of my code for clearity on the utility function.

I hope you can shed some light on this matter. Thank you very much!

Greetings, Mariska

--
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/6eee3a9e-7c2b-4dca-9933-5ee0889b4fafn%40googlegroups.com.
<Biogeme part of code.py><size function.png>

--
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/ae2c56b8-954e-4d44-a174-0174176ccc78n%40googlegroups.com.