Date:         Wed, 18 Oct 2006 20:02:44 -0400
Reply-To:     Peter Flom <flom@NDRI.ORG>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Peter Flom <flom@NDRI.ORG>
Subject:      Re: What's the difference between parametric and non-parametric
Comments: To: tw2@MAIL.COM
Content-Type: text/plain; charset=US-ASCII

>>> Tom White <tw2@MAIL.COM> 10/18/06 5:21 PM >>> wrote
<<<
I am new to this list. I am an inexperienced statistical modeler, just beginning my work now.
>>>
Welcome!  We all started somewhere.

<<<
I will be posting lots of questions on this list...SAS and statistically related.
>>>
You'll probably get lots of answers, too.  Hopefully they won't contradict each other!

<<<
I would like to offer you my apologies if I offend some of you with my simple SAS & STAT questions. I will learn eventually from you all and I may even begin to ask difficult questions later on.
>>>>

Hard to offend with questions....


<<<
So, here is my first question:

When we are talking about parametric vs non-parametric approaches, when it comes to statistical formulas, modeling, etc., what are we talking about?
>>>

A big topic.  The simple answer is that parametric models have parameters, and nonparametric ones don't.  But that's not helpful

Here's one example that may be more so:

Regression.  One form of parametric regression is ordinary least squares regression.  Here, the dependent variable is a linear combination of the independent variables.  You estimate the parameters, which are the coefficients that you multiply the independent variables' values by

So

Y = a + b1X1 + b2X2 + b3X3 + e


a, b1, b2, and b3 are parameters

all well and good. Easy to solve, easy to interpret. But this model makes assumptions. (let us know if you want a review).

in nonparemtric regression, the DV is any function of the IVs.
What form the function takes varies.  One nonparamteric regression technique is loess regression.
Another is restricted cubic splines (although this might be called semi-parametric, since the cubic splines do have parameters)

Does this help at all?

Too simple? Too advanced?

Write back to SAS-L (not just me) and let us know


Peter