Non-Linear Regression and Dummy Variables

[JohnYork]

Hi,

I'm new to EViews and this forum. Hopefully, I've posted this on the right board.

I am really struggling with a number of regressions I need to carry out. I would really appreciate any advice or help.

My Dependent Variable is Delay (D) and my Explantory Variable is Capacity Use (Cap). The idea being that as Capacity Use increases so does the level of Delay.

I have a data set with 256 Observations. These are for 1 Route which is divided into 16 sections. Each section has 16 different observations (which equate to different times of the day). Note, some of my Delay observations are 0 so I can't use logs.

I have a number of equations that I need to test.

A Linear one :-

D=a+(b*Cap) where a is a section coefficient and b is a route coefficient.

I have been able to run this in eviews using the input.

D c Cap @expand(section, @dropfirst)

with 'section' in the data set being an id code showing which section an observation belongs to.

This gives me individual values and results for each of the 16 sections ('a's) and an overall value and result for the 1 overall route (the 'b').

However, I also have 2 non-linear equations and E Views returns an error when I try and use @expand with a non-linear equation.

My equations are:-

D=a*exp(b*Cap) and D=a*Cap^b

So How do I do what I want to do in EViews (its EViews 7 by the way)?

Any advice on how to tackle this would be greatly appreciated.

Regards
John

[EViews Gareth]

It sounds like you should look into using the Pool object. I'd suggest reading up on it in the user manual, then coming back with any other questions.

[JohnYork]

Hi,

Thanks for the quick reply.

Pool Object does look to be a way forward. I'm busy sorting out my data into a suitable format to give it a go.

However, I don't see how you input a Non-Linear equation. The examples in the User Manual all seem to be based on linear equations.

I haven't got access to the Software where I am now but I will have a go tomorrow.

Is there an option for entering Non-Linear equations using Pooled Data ? e.g. my exponential equation D=a*exp(b*Cap).

Regards
John

[EViews Gareth]

Ah, indeed, I'm an idiot. You cannot estimate non-linear in a pool, sorry.

You might just have to write out the non-linear by hand in a system or something.

[JohnYork]

Hi,

I tried running my equations using ''System Object" in EViews this morning.

It worked like a dream. It allowed me to do the analysis exactly as I needed to and gave me the type of results I was looking for. I still have 1 or 2 things to sort out but I am now well on the way with my work.

Thanks enormously for your help.

Regards

John :D