0
$\begingroup$

I would like to check for differences in growth rate between groups. I have three main groups miRs and for each group I have a treatment and a Neg. I want to compare treatment vs Neg for all groups. Could someone have a look at my setup? Why is the variable mir-135b-5p nor showing up in the results?

> lm <- lme(weight~time*cond, random=~time|miRs, data=testDose)
> anova(lm)
            numDF denDF   F-value p-value
(Intercept)     1    38 233748.85  <.0001
time            1    38    398.12  <.0001
cond            3    38      7.14  0.0006
time:cond       3    38      2.34  0.0887
> summary(lm)
Linear mixed-effects model fit by REML
 Data: testDose 
       AIC      BIC    logLik
  51.94759 72.21414 -13.97379

Random effects:
 Formula: ~time | miRs
 Structure: General positive-definite, Log-Cholesky parametrization
            StdDev       Corr  
(Intercept) 3.279719e-06 (Intr)
time        9.458478e-10 0     
Residual    1.951541e-01       

Fixed effects: weight ~ time * cond 
                        Value  Std.Error DF  t-value p-value
(Intercept)         12.322245 0.16900842 38 72.90906  0.0000
time                 0.021713 0.00257138 38  8.44398  0.0000
condmir-21-3p        0.048565 0.23901400 38  0.20319  0.8401
condmir-584-5p       0.125029 0.23901400 38  0.52310  0.6039
condNeg              0.021297 0.19515412 38  0.10913  0.9137
time:condmir-21-3p  -0.002460 0.00363648 38 -0.67658  0.5028
time:condmir-584-5p -0.006198 0.00363648 38 -1.70431  0.0965
time:condNeg         0.001352 0.00296918 38  0.45549  0.6513
 Correlation: 
                    (Intr) time   c-21-3 c-584- condNg t:-21- t:-584
time                -0.913                                          
condmir-21-3p       -0.707  0.645                                   
condmir-584-5p      -0.707  0.645  0.500                            
condNeg             -0.866  0.791  0.612  0.612                     
time:condmir-21-3p   0.645 -0.707 -0.913 -0.456 -0.559              
time:condmir-584-5p  0.645 -0.707 -0.456 -0.913 -0.559  0.500       
time:condNeg         0.791 -0.866 -0.559 -0.559 -0.913  0.612  0.612

Standardized Within-Group Residuals:
       Min         Q1        Med         Q3        Max 
-1.4935997 -0.5794501 -0.2101814  0.4779404  1.4872837 

Number of Observations: 48
Number of Groups: 3  

myData

> testDose
     weight time        cond        miRs
1  13.10760   24   mir-21-3p   mir-21-3p
2  13.21659   48   mir-21-3p   mir-21-3p
3  13.59320   72   mir-21-3p   mir-21-3p
4  14.18753   96   mir-21-3p   mir-21-3p
5  13.16919   24         Neg   mir-21-3p
6  13.51846   48         Neg   mir-21-3p
7  14.05866   72         Neg   mir-21-3p
8  14.84804   96         Neg   mir-21-3p
9  12.79820   24   mir-21-3p   mir-21-3p
10 13.12730   48   mir-21-3p   mir-21-3p
11 13.69247   72   mir-21-3p   mir-21-3p
12 14.48472   96   mir-21-3p   mir-21-3p
13 12.81832   24         Neg   mir-21-3p
14 13.17344   48         Neg   mir-21-3p
15 13.78877   72         Neg   mir-21-3p
16 14.44471   96         Neg   mir-21-3p
17 13.11672   24 mir-135b-5p mir-135b-5p
18 13.31708   48 mir-135b-5p mir-135b-5p
19 13.78559   72 mir-135b-5p mir-135b-5p
20 14.54087   96 mir-135b-5p mir-135b-5p
21 13.16919   24         Neg mir-135b-5p
22 13.51846   48         Neg mir-135b-5p
23 14.05866   72         Neg mir-135b-5p
24 14.84804   96         Neg mir-135b-5p
25 12.79218   24 mir-135b-5p mir-135b-5p
26 13.18126   48 mir-135b-5p mir-135b-5p
27 13.78006   72 mir-135b-5p mir-135b-5p
28 14.48629   96 mir-135b-5p mir-135b-5p
29 12.81832   24         Neg mir-135b-5p
30 13.17344   48         Neg mir-135b-5p
31 13.78877   72         Neg mir-135b-5p
32 14.44471   96         Neg mir-135b-5p
33 13.06603   24  mir-584-5p  mir-584-5p
34 13.10624   48  mir-584-5p  mir-584-5p
35 13.27287   72  mir-584-5p  mir-584-5p
36 13.96981   96  mir-584-5p  mir-584-5p
37 13.16919   24         Neg  mir-584-5p
38 13.51846   48         Neg  mir-584-5p
39 14.05866   72         Neg  mir-584-5p
40 14.84804   96         Neg  mir-584-5p
41 12.81925   24  mir-584-5p  mir-584-5p
42 13.03732   48  mir-584-5p  mir-584-5p
43 13.59867   72  mir-584-5p  mir-584-5p
44 14.15521   96  mir-584-5p  mir-584-5p
45 12.81832   24         Neg  mir-584-5p
46 13.17344   48         Neg  mir-584-5p
47 13.78877   72         Neg  mir-584-5p
48 14.44471   96         Neg  mir-584-5p
$\endgroup$
1
  • $\begingroup$ "I want to compare treatment vs Neg for all groups." I don't understand why you use a mixed effects model then. Just do separate tests for each group and adjust the p-values for multiple testing. $\endgroup$
    – Roland
    May 19, 2016 at 9:31

1 Answer 1

0
$\begingroup$

Your cond is a factor with 4 values, these are modeled by 3 binary indicator variables, the fourth value corresponds to the 3 binary variables being 0. The baseline is the fourth value and is already accounted for in the intercept and time only the deviation from these is expressed for the others.

$\endgroup$
4
  • $\begingroup$ But how do I get the p-value for mir-135b-5p $\endgroup$ May 18, 2016 at 15:22
  • $\begingroup$ Sorry, can't help you with that. $\endgroup$
    – bnord
    May 18, 2016 at 15:28
  • 1
    $\begingroup$ @user2300940 mir-135b-5p is collapsed into the intercept, because otherwise the covariance matrix is singular. You have a treatment contrast by default in R, so all the tests you make are comparing to that baseline, i.e. mir-135b-5p, because that is the first level in the factor. You probably want to reorder the levels in the factor, such that Neg is the baseline and compare to that, but I am not sure if that solves your problem. The question you are trying to answer sounds more like anova. $\endgroup$
    – Gumeo
    May 18, 2016 at 15:51
  • $\begingroup$ So maybe put Neg first since its similar for all observations, then put all the different observations (miRs) after? $\endgroup$ May 19, 2016 at 6:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.