[R] problems understanding error term in aov()

[Jonas Mandel]

Hi all,

I'm having some trouble in understanding how to ste the Error() term in  
the aov() function when fitting a hierarchical ANOVA.
I have data concerning the expression of 2 miRNAs in 3 different cell 
lines, with 2 different extraction methods. The data is organized as 
follows :
    Line    Extraction    Target    Expression
1    BC54    miRNA    RNU48    22.48
2    BC54    miRNA    RNU48    22.36
3    BC54    miRNA    221    26.49
4    BC54    miRNA    221    26.42
5    BC54    RNAtot    RNU48    22.95
6    BC54    RNAtot    RNU48    22.98
7    BC54    RNAtot    221    26.35
8    BC54    RNAtot    221    26.43
9    BC108    miRNA    RNU48    24.19
10    BC108    miRNA    RNU48    24.01
11    BC108    miRNA    221    29.94
12    BC108    miRNA    221    29.61
13    BC108    RNAtot    RNU48    22.42
14    BC108    RNAtot    RNU48    22.29
15    BC108    RNAtot    221    26.79
16    BC108    RNAtot    221    26.73
17    BC233    miRNA    RNU48    21.84
18    BC233    miRNA    RNU48    21.69
19    BC233    miRNA    221    25.95
20    BC233    miRNA    221    25.98
21    BC233    RNAtot    RNU48    22.55
22    BC233    RNAtot    RNU48    22.58
23    BC233    RNAtot    221    25.11
24    BC233    RNAtot    221    25.45

For the two miRNA and for each cell line I want to compare the 
expression value between the two extraction methods. As there are no 
random effects I tried to use the aov() function to do so.  I don't know 
which of the two following models is the "good" one :

This one :
 > mir221.aov=aov(Ct~Extraction+Error(Target/Line/Extraction),data=mir221)

with this model the extraction has no significant effect :
 > summary(mir221.aov)
Error: Target
          Df Sum Sq Mean Sq F value Pr(>F)
Residuals  1 99.675  99.675              

Error: Target:Line
          Df  Sum Sq Mean Sq F value Pr(>F)
Residuals  4 16.9779  4.2445              

Error: Target:Line:Extraction
           Df  Sum Sq Mean Sq F value Pr(>F)
Extraction  1  2.8912  2.8912  1.3568 0.2966
Residuals   5 10.6545  2.1309              

Error: Within
          Df   Sum Sq  Mean Sq F value Pr(>F)
Residuals 12 0.164150 0.013679

or this one :
 > mir221.aov=aov(Ct~Extraction+Error(Target/Line),data=mir221)

with this model the extraction has a significant effect:
 > summary(mir221.lm)
Error: Target
          Df Sum Sq Mean Sq F value Pr(>F)
Residuals  1 99.675  99.675              

Error: Target:Line
          Df  Sum Sq Mean Sq F value Pr(>F)
Residuals  4 16.9779  4.2445              

Error: Within
           Df  Sum Sq Mean Sq F value  Pr(>F) 
Extraction  1  2.8912  2.8912  4.5431 0.04794 *
Residuals  17 10.8187  0.6364    

Could you please help me understanding which is the right model for my 
question ? Thanks by advance

Jonas Mandel
U900 - Bioinformatics unit
Institut Curie
Paris