Scale parameter

[Nizam Husen]

Hi Prof,

I am running a model and I defined the scale parameter as follows:

Scale = Beta('Scale',1.45608965291534,-1,1.5,0,'Scale' )

When it converges, I get the following: 

When I fix the scale to 1, this problem disappears.

Regards

Nizam

[Bierlaire Michel]

In discrete choice, the scale is not identified.

Moreover, it cannot be negative. So the lower bound set at -1 is incorrect. 

On 21 Oct 2021, at 09:51, Nizam Husen <abba...@gmail.com> wrote:

Hi Prof,

I am running a model and I defined the scale parameter as follows:

Scale = Beta('Scale',1.45608965291534,-1,1.5,0,'Scale' )

When it converges, I get the following: 

<image.png>

When I fix the scale to 1, this problem disappears.

Regards

Nizam

--
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/CANdHA5Am%2BhX7oXkYqZ%3DKMUVfjiS_n22Ayg0W_sHbbDKmXvZrLg%40mail.gmail.com.

[Nizam Husen]

Dear Prof,

Apologies for reposting this question and thanks for your suggestions about this couple of days ago.

You suggested that the lower bound of the scale parameter should not be negative, which I did wrong at first. 

I tries with other non-negative lower bound values, but still, the model is unidentified.

I defined the scale parameter as:

Scale = Beta('Scale',1.45608965291534, 0,1.5,0,'Scale' ) and other non-negative values. 

An example of utility function:

V1 = exp(Scale*LV_VID)*(ASC + Plantation*A1C1SFW1 + Repurposed*A1C1SFW2 + Farm*A1C1SFW3 + Poor*A1C1LFW1 + Jobs*A1C1LFW2 +  Wildlife*A1C1LFW3  + Distance*A1C1DFW + EC       + Cost*A1C1CFW)

where LV_VID is a latent variable. 

I am grateful if you have some suggestions for me.

Kind regards,