[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
RE: st: Nonlinear regression and constraints
> -----Original Message-----
> From: email@example.com
> [mailto:firstname.lastname@example.org] On Behalf Of
> Richard Williams
> Sent: Tuesday, June 28, 2005 10:26 PM
> To: email@example.com
> Subject: Re: st: Nonlinear regression and constraints
> At 09:04 PM 6/28/2005 -0700, Daniel Schneider wrote:
> >constraint define 1 X1MX1SQDIVX3 >= 0
> >constraint define 2 X2SQMX2DIVX3 >= 0
> >constraint define 3 X1MX1SQDIVX3 <= 1
> >constraint define 4 X2SQMX2DIVX3 <= 1
> >Unfortunately, constraints() seems not to work with -nl-?
> >Is there any other way to to do the constraints with -nl-?
> Even if -nl- allowed the constraints option, I don't think
> these would be
> legal constraints; as far as I know, you can't use something
> like >=, you
> can only use =.
That might be possible. Let me clarify what I want to do: I want to find
the best (i.e. with the lowest SS) parameter values which are between 0
and 1 (including both), because by definition they can only be between
those two values.
> Also, my impression is that -nl- doesn't need the constraints option
> because constraints can be specified using the -nl- command itself.
Can you tell me how that can be done?
> I'd be curious to know if there is some direct way to specify
> a range of
> values for the constraint, like you want; as far as I know,
> there always
> has to be some sort of equality statement.
Perhaps I am wrong thinking here about constraints. Perhaps "range of
values" would be a better term, as I explained above. My -nl- commands
gives me reasonable results in a statistical sense, but I know that by
definition my alpha and beta cannot be larger 1 or smaller than 0.
> Is there some way to transform the equation, so that, say,
> you wind up
> taking the inverse logit of the parameters of the transformed
> equation to
> get back to the parameters of the original equation? The
> inverse logit of
> any number will always range between 0 and 1. You'll see programs do
> little tricks like estimate the log of a parameter when the parameter
> itself needs to be positive.
That is of course and interesting idea. I'll have to check that and
reconsider the whole equation...
* For searches and help try: