Skip to content

Help in using multcomp.

5 messages · A Ezhil, Chuck Cleland

Original
A Ezhil
# 
Hi All,

I am trying use 'multcomp' for multiple comparisons
after my ANOVA analysis. I have used the following
code to do ANOVA:

dat <- matrix(rnorm(45), nrow=5, ncol=9)
f <- gl(3,3,9, label=c("C", "Tl", "T2"))

aof <- function(x) {
        m <- data.frame(f, x);
        aov(x ~ f, m)
}
amod <- apply(dat,1,aof)

Now, how can I use 'glht' for the above amod. I know
that I cannot use simply 

glht(amod, linfct = mcp(f = "Dunnett")). 

Also, if I want to use Dunnett for comparing C vs (T1
and T2), how can I specify this in the glht function.

Thanks in advance. 
Regards,
Ezhil
Chuck Cleland
#
A Ezhil wrote:
Since amod is a list of models rather than one model, do you want
something like this?

lapply(amod, function(x){summary(glht(x, linfct = mcp(f = "Dunnett")))})
How about doing it with user-defined contrasts?

contr <- rbind("C - T1   " = c(-1, 1, 0),
               "C - T2   " = c(-1, 0, 1),
               "C - All T" = c(-1,.5,.5))

lapply(amod, function(x){summary(glht(x, linfct = mcp(f = contr)))})

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      A Ezhil
    

    

      
      #
    

  


  

      
Hi Chuck,

Thank you very for this help. I am able to store the
results. Now, I am facing the following problems:

1. I am trying to extract only the p-values from 

dunres <- lapply(amod, function(x){summary(glht(x,
linfct=mcp(f = contr)))})

but I am stuck. If I use dunres[[1]], it displays the
results. But I don't know how to extract the p value
from this. 

2. If I want to get raw pvalues instead of adjusted
ones, what should I do in summary(glht)? 

Thanks again for your help. I look forward to your
reply.

Kind regards,
Ezhil

        
--- Chuck Cleland <ccleland at optonline.net> wrote:

        

            

      
      
      
    

        

      
      
    

        

      
      
    

        
____________________________________________________________________________________
Need a quick answer? Get one in minutes from people who know.
Ask your question on www.Answers.yahoo.com

  
  


  


      

    

    

      
      

        


  

        

      Chuck Cleland
    

    

      
      #
    

  


  

      
A Ezhil wrote:

            

      
      
      
    

        
Ezhil:

str(summary(glht(amod[[1]], linfct=mcp(f = contr)))) suggests a way to
extract just the p values.  Try this:

library(multcomp)

dat <- matrix(rnorm(45), nrow=5, ncol=9)
f <- gl(3,3,9, label=c("C", "Tl", "T2"))

aof <- function(x) {
        m <- data.frame(f, x);
        aov(x ~ f, m)
}

amod <- apply(dat,1,aof)

my.pvals <- sapply(amod, function(x){summary(glht(x, linfct=mcp(f =
contr)))$test$pvalues})

rownames(my.pvals) <- rownames(contr)
colnames(my.pvals) <- paste("amod", 1:5, sep="")

my.pvals
               amod1     amod2     amod3     amod4     amod5
C - T1    0.03348242 0.3581771 0.9873633 0.9764219 0.9225445
C - T2    0.67794496 0.7138491 0.2183949 0.9962458 0.8439224
C - All T 0.10621039 0.4344881 0.4913578 0.9970689 0.8519888

            

      
      
      
    

        
See the test argument of summary.glht().  You could do something like
this:

my.pvals <- sapply(amod, function(x){summary(glht(x, linfct=mcp(f =
contr)), test = adjusted("none"))$test$pvalues})

rownames(my.pvals) <- rownames(contr)
colnames(my.pvals) <- paste("amod", 1:5, sep="")

my.pvals
               amod1     amod2     amod3     amod4     amod5
C - T1    0.01770489 0.2197886 0.8985143 0.8611198 0.7466660
C - T2    0.46651736 0.4994115 0.1278132 0.9448342 0.6364631
C - All T 0.05953715 0.2733017 0.3149223 0.9512189 0.6464735

hope this helps,

Chuck Cleland

            

      
      
      
    

        

      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      A Ezhil
    

    

      
      #
    

  


  

      
Hi Chuck,

Thank you very much for this is help. It is fantastic
and the code works perfectly.

Thanks again, Chuck.

Kind regards,
Ezhil

        
--- Chuck Cleland <ccleland at optonline.net> wrote:

        

            

      
      
      
    

        

      
      
    

        

      
      
    

            

      
      
      
    

        
____________________________________________________________________________________

            

      
      
      
    

        
____________________________________________________________________________________
Get your own web address.