Re: Regression Basics

The linear regression model assumes that the parameters remain constant.
The estimated parameters are a function of observation and hence you have
some levels of uncertainty.

<NPDave@xxxxxxxxx> wrote in message
news:1169661306.798366.325980@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
I'm trying to figure out some simple regression theory and would
appreciate help/comments from the talented folks reading this.

In Simple Linear Regression (SLR),

Yi=beta0 + beta1 Xi + epsilon i

we assume that the parameters beta0 and beta1 are constants along with
Xi

Now, the estimated SLR eqn is :

Y^i = bo + b1 Xi , where "^" denotes hat which is an estimate of the
parameter

Now, when we make inferences for bo and b1, we talk about their
expectations, variances and sampling distributions which means that we
are saying at the outset that they are random variables (RVs).

So, then would it be correct to say that bo, b1 are RVs, but the
paramaters beta0, beta1 are NOT (they are constants) ?

If true, then isn't it rather strange that the parameters are NOT RVs,
but the
estimates (b0, b1) of the parameters ARE RVs ? Is there an intuitive
reasoning ?

Thanks,
Dave

P.S: I havesome follow up questions and will be putting them up here
shortly



.

Relevant Pages

  • Re: Statistics Basics
    ... In Simple Linear Regression, ... we assume that the parameters beta0 and beta1 are constants along with ... the estimated SLR eqn is: ... If true, then isn't it rather strange that the parameters are NOT RVs, ...
    (sci.math)
  • Assumptions in Linear Regression
    ... I have to show that in a linear regression model the following ... inferences and not that the assumptions somehow ... imply linear regression. ...
    (sci.stat.math)
  • assessing multicollinearity in logistic regression
    ... The first message did not appear to go through. ... In linear regression, one can look at tolerance/VIF. ... linear regression model and use the VIFs from that? ...
    (sci.stat.edu)
Quantcast