R help - Oct 2008 - interpreting Shapiro-Wilks test result

Hi all,
   I am newbie in using R software and also doing statistical test. I want to
know if my data in in normal distribution. I have 2 groups of data and I did
calculate Shapiro Wilks using R software. Here is the results:

Group 1: W = 0.9206, p-value = 0.01683
Group 2: W = 0.9626, p-value = 0.4694

I am not quite sure what default confidence level (CF) is used in calculating
Shapiro Wilks. Else, may I choose my own CF. Let say, if I choose CF = 0.01,
therefore I may approve that Group 1 and Group 2 data are normal. If not, what
should I do. My next plan is to apply T Test for both groups (assuming both are
normal). I really need advice from experts here.

Cheers,
Halizah



      
	[[alternative HTML version deleted]]

Em Qua, 2008-10-08 ?s 17:41 -0700, Halizah Basiron
escreveu:
> Hi all,
>    I am newbie in using R software and also doing statistical test. I want
to know if my data in in normal distribution. I have 2 groups of data and I did
calculate Shapiro Wilks using R software. Here is the results:
> 
> Group 1: W = 0.9206, p-value = 0.01683
> Group 2: W = 0.9626, p-value = 0.4694
> 
> I am not quite sure what default confidence level (CF) is used in
calculating Shapiro Wilks. Else, may I choose my own CF. Let say, if I choose CF
= 0.01, therefore I may approve that Group 1 and Group 2 data are normal. If
not, what should I do. My next plan is to apply T Test for both groups (assuming
both are normal). I really need advice from experts here.
> 
> Cheers,
> Halizah
Hi Halizah,

In my opinion don't exist confidence level of test. What exist is
confidence level of your experiment.

If you plan your experiment with alpha probability you have
100*(1-alpha) confidence level and you p-value need minor than alpha to
named "significant"

So if you need 99% of significance the two groups is normal and you
using t.test. 


-- 
Bernardo Rangel Tura, M.D,MPH,Ph.D
National Institute of Cardiology
Brazil