1st order or 2nd order distribution

[bradyj7]

Hi,

I'm fitting a distribution to the starting times of the first car journey in the day. I have a sample of 3,000 journey starting times. I am assuming that this sample represents the population well.

I'm fitting a non parametric distribution.

But my question is, should I fit a 1st order or 2nd order distribution?

It should be second order correct? because there is uncertainty about the parameters?

Thanks

 

[chiro]

Hey bradyj7.

What are you saying doesn't make sense because when you fit something to something else, that something else has existing form that you fit to.

What exactly are you trying to fit to?

 

[bradyj7]

Hello Chiro,

Well I originally fitted a parametric distribution to the data and a Johnson unbounded distribution was the best fit using an SIC score.

Here it is:

https://dl.dropbox.com/u/54057365/All/g1.JPG

But I did not think that the fit was good enough, so I fitted an Ogive distribution, a non parametric 1st order distribution.

Here it is
https://dl.dropbox.com/u/54057365/All/g2.JPG


However because I only have a sample of journey start times, I am thinking that there is uncertainty about the non-parametric distribution, so perhaps I should fit a non parametric 2nd order distribution. The red line is the uncertainty distribution.

here it is:
https://dl.dropbox.com/u/54057365/All/g3.JPG



What do you think?

Thanks