Date: Nov 27, 2012 11:32 PM
Author: Ray Koopman
Subject: Re: Here's how I did logs ...

On Nov 27, 4:06 pm, djh <halitsk...@att.net> wrote:
> You wrote:
>
> ***
> In your previous post, you said
> "[the] underlying regression is c on (u',u'^2), where u' = u/(1+u)."
>
> Now are you saying that the underlying regression was not that,
> but was ln(c) on (u',u'^2), where u' = ln(u/(1+u^2)) ?
> ****
>
> Sorry for the earlier reference without logs - at the time I typed it
> I was assuming logs to be understood.
>
> So the way I did logs was as follows;
>
> c on (u',u'^2)
>
> where u' = log( $u / ( 1 + $u ) )
> u'^2 = ( log( $u / ( 1 + $u ) ) )^2
>
> (I was going "by analogy", i.e. to square u', whatever it is. )
>
> So, please confirm if:
>
> 1) you want logs or not
> 2) if so, is the above correct? (If not, then what?)


No logs, please. Make it c on (u',u'^2)

where u' = $u / ( 1 + $u )
u'^2 = ( $u / ( 1 + $u ) )^2 .