Prev: Resized Command Bar
Next: question on first events: pasted data layout and code
From: juliaw on 2 Aug 2010 17:06
I am almost ashamed of this post as I try not to delve into statistics
without some theoretical background of what I am doing but since I
have gone there here is my question:

I am using PROC GLM to predict birthweight using the variables race
marital status site and study group (my main variable of interest). I
ran my model but also used a few estimate statement to compare a the
study groups of interest:

For example: ESTIMATE 'Compare 3rd & 1st' group -1 0 1 0

This seems like an easier alternative to re-running models and
changing the reference group for my group variable each time.

My question is when I am writing up my formula for my general linear
model, how do I include the estimate statement into my formula / how
do I discuss it appropriately? Is it even necessary?

Also, on a side note, how does the estimate statement differ from the
contrast statement?

Thanks!

From: Paige Miller on 3 Aug 2010 12:18
On Aug 2, 5:06 pm, juliaw <julia.wen...(a)live.com> wrote:
> I am almost ashamed of this post as I try not to delve into statistics
> without some theoretical background of what I am doing but since I
> have gone there here is my question:
>
> I am using PROC GLM to predict birthweight using the variables race
> marital status site and study group (my main variable of interest). I
> ran my model but also used a few estimate statement to compare a the
> study groups of interest:
>
> For example: ESTIMATE 'Compare 3rd & 1st' group -1 0 1 0
>
> This seems like an easier alternative to re-running models and
> changing the reference group for my group variable each time.
>
> My question is when I am writing up my formula for my general linear
> model, how do I include the estimate statement into my formula / how
> do I discuss it appropriately? Is it even necessary?
>
> Also, on a side note, how does the estimate statement differ from the
> contrast statement?


I don't think you have to include it into your formula. You can simply
say the difference in means between the first and third groups is
YYY.YY units, and this is (or is not) statistically significant.

The CONTRAST statement tests hypothesis that can be expressed as
contrasts of your group means. The ESTIMATE statement estimates the
value of a contrast and also tests its significance.

--
Paige Miller
paige\dot\miller \at\ kodak\dot\com
 | 
Pages: 1
Prev: Resized Command Bar
Next: question on first events: pasted data layout and code