Date: Tue, 12 Dec 2006 16:22:42 -0500 Sigurd Hermansen "SAS(r) Discussion" Sigurd Hermansen Re: Stats SAS Project for college statistics course, issue with results To: Romel Lira <8e6c604a0612121259p39d0b387k5c0f7f9bd4e5f87a@mail.gmail.com> text/plain; charset="us-ascii"

Romel: If you think of a believable story as prior knowledge that holds you back from making a false leap based on very limited data, then you will be OK. In the real world it works better if you start with a believable story about what determines what and then use data to make a case for that story or reject it. In any situation you'll need enough statistical power to reject or accept a model. With enough statistical power, you will be able to validate a model using data not used in the original model estimation, and then be able to evaluate prediction errors of a model in another data subset not used to estimate the model. I agree that the group variable seemed more a prediction than a predictor. As for exponential smoothing, you may find it more instructive to start simpler transforms of predictors (y-f(x,x**2,....)). I had to line up columns in the CARDS statement to read data copied from your email, but I would hope that you have printed your SAS dataset and check to make sure that program reads data correctly. I also substituted a period '.' for a '*' because SAS treats a period in numeric input as a missing value. The procedures appeared to work fine. Sig

-----Original Message----- From: Romel Lira [mailto:rlira007@gmail.com] Sent: Tuesday, December 12, 2006 3:59 PM To: Sigurd Hermansen Cc: sas-l@listserv.uga.edu Subject: Re: Stats SAS Project for college statistics course, issue with results So in other words, if I can tell a believable story of how the model I have fits the data, I should be ok? I decided to remove the group designation and use that info (group/GNP) for a separate calculation to demonstrate that there is a significant difference between groups in terms of GNP. Mainly because of the very reasons you state, I wasn't sure if it was possible to throw both types of data into the same modeling, but I find it telling that the most significant data from the ANOVA print out was the group number.

My main fear is not that the data is wrong, but that my SAS programming is incorrect. Should I use exponential smoothing on the GNP values? Thanks for taking the time to write your response, you've certainly provided a new outlook on my 2nd semester of statistics. Romel On 12/12/06, Sigurd Hermansen <HERMANS1@westat.com> wrote: