Re: extreme values



thierry wrote:
> hi,
> I'm working with a regression equation like : Y = - 2000 + 544*var1 +
> 1166*var2 - 487*var3.
> as you see, I have an intercept (-2000) negative, and it is due to extreme
> values on my data.
> what should I do ? ideally, I want to have a positive intercept. !!!!!!!!!
> Thank you.

It sounds like you may be using a procedure whose assumptions are
unmet; I take it these are what are often called "influential
outliers"?

Is the linear model really applicable over the range of the data (did
you check diagnostics at all)?

If the model does in fact seem linear, you may wish to revisit the
assumptions of equal variance and normal errors, and perhaps consider a
procedure more robust to influential outliers.

Why do you "want" a positive intercept? I presume there's some a priori
reason you believe the expected response should be positive over some
region? Such information can be incorporated in the model (in various
ways).

Glen

.

Relevant Pages

  • extreme values
    ... as you see, I have an intercept negative, and it is due to extreme ... I want to have a positive intercept. ... Prev by Date: ...
    (sci.stat.edu)
  • extreme value
    ... as you see, I have an intercept negative, and it is due to extreme ... I want to have a positive intercept. ... Prev by Date: ...
    (sci.stat.consult)
  • Re: multiple regression (intercept)
    ... One can fit a model without an intercept, ... Neither of course will the linear model without intercept ... I interpreted the OP's question as one where he had noticed something that seemed silly, but there was nothing in his comments to suggest that his data was anywhere near this region of the parameter space. ...
    (sci.stat.math)
  • Re: multiple regression (intercept)
    ... > illywhacker wrote: ... >>>I ask this question because I have a negative intercept in my equation (Y ... range of values where the behaviour is nonlinear. ... Neither of course will the linear model without intercept ...
    (sci.stat.math)