Skip to content

Calculating the t-test for each row

5 messages · Keizer_71, Henrique Dallazuanna, John Kane +2 more

Keizer_71
# 
Hi Everyone,

I need some simple help.

Here are my codes

##########will give me 10000 probesets####################
data.sub = data.matrix[order(variableprobe,decreasing=TRUE),][1:10000,]
dim(data.sub)
data_output<-write.table(data.sub, file = "c://data_output.csv", sep = ",",
col.names = NA) 

When i export to excel, it shows me this. This is just a short version.
There are 1000 rows and 140 columns

	Sample_1_D	Sample_1_C	Sample_2_D	Sample_2_C
1	2.425509867	11.34031409	11.46868531	11.75741478


Here is my question: How do create a new row and calculate the t-test so
that it will give me the p-value

Here is what i am looking for. The p-value is not correct but just an
example. It needs to calculate the entire each row. There are 10000 rows and
140 columns.

thanks
Kei

	Sample_1_D	Sample_1_C	Sample_2_D	Sample_2_C    p-value
1	2.425509867	11.34031409	11.46868531	11.75741478     .00341111

I tried something like this.

t.test(data.sub,mu=0)

I am pretty new to R. I think it is showing me the entire p-value.
Henrique Dallazuanna
# 
You can try this:

cbind(data.sub, p.value=apply(data.sub, 1, function(x)t.test(x)$p.value))
On 03/03/2008, Keizer_71 <christophe.lo at gmail.com> wrote:

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      John Kane
    

    

      
      #
    

  


  

      
If I understand you correctly what you want to do is
do t-test (mu=0) for each column of the data. 

Treating the data as a data.frame rather than a matrix
you can do something like this and then pick out the
p-values but with 140 t-tests I don't know what you'll
get in terms of anything meaninful.  

======================================================


aa <- data.frame(a=rnorm(25, 5, 2), b=rnorm(1:25,
0,1))
mytea <- apply(aa, 2, t.test)

tresults <- lapply(mytea, function(.tres) {      
 
data.frame(t.value=.tres[1],dfs=.tres[2],conf.int1=.tres$conf.int[1],conf.int2=
 .tres$conf.int[2],p.value=.tres[3])
      })

finalresults <- do.call(rbind, tresults) 

=====================================================
(Thanks to Mark Leeds for the lapply approach)

        
--- Keizer_71 <christophe.lo at gmail.com> wrote:

        

            

      
      
      
    

        
data.matrix[order(variableprobe,decreasing=TRUE),][1:10000,]

            

      
      
      
    

        
http://www.nabble.com/Calculating-the-t-test-for-each-row-tp15808716p15808716.html

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Benilton Carvalho
    

    

      
      #
    

  


  

      
apparently you want to check the genefilter package...

it defines functions like:

rowttests
colttests
rowFtests
colFtests
rowVars
rowSds

moreover, a quick look at Biobase is recommended...

that would save you lots of time as you wouldn't have to reinvent the  
wheel.

b

        
On Mar 3, 2008, at 12:42 PM, Henrique Dallazuanna wrote:

        

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      James W. MacDonald
    

    

      
      #
    

  


  

      
Using t.test() in this case will likely be very slow. A faster 
alternative would be to use rowttests() from the genefilter package of 
Bioconductor.

Best,

Jim

        
Henrique Dallazuanna wrote: