[BioC] reflexion on comparaison of normalization

[gregory voisin]

Hello Bioconductoriens
 
I begin my studies of analysis of microarray..
I would like to want your opinion on my idea:
To use the better normalization for my design , i practise all the normalization to compare them 
1. I looked at the dispersion and position parameter for my no-normalized data: mean, max min, var, skewness, kurtosis: goal: to know the degree of normality for my data(file .cel treated by bioconductor)and after compared this results.
 
2.I normalize : RMA ,QUANTILE , MAS5.and i calcule  the statictic parameters.
My surprise : the different parameters haven' tgot the same scale:
example;
mean for  no normalized data 

median

96.0

74.0

81.0

81.0

mean for  RMA normalized data

mean

5.294393

5.302782

5.296402

5.290933
 meanfor quantile normalized data

mean

219.2156

216.6649

218.4102

217.4639

why the difference between this values?
 
 
3.I think to verify these differents normalizations with kurtosis ans skewness parameter and a normality test..
what do you think of this??
 
4. The last question, for this comparaison I use only four controls of my design (four biologicals replicats)...  can I think that the best normalization fot these 4 controls is too the best normalization for my global design?
 
thanks
 
greg
bioinformaticien 
CHUM-montreal
 


		
---------------------------------




és pour dialoguer instantanément avec vos amis.Téléchargez GRATUITEMENT ici !
	[[alternative HTML version deleted]]