[BioC] understanding ACME

Ramon Diaz-Uriarte rdiaz at cnio.es
Mon Feb 4 19:50:27 CET 2008


hits=-2.6 testsºYES_00
X-USF-Spam-Flag: NO

Dear All,

I am trying to understand how the ACME package (by Sean Davis) works, but I 
think there is something I am missing about the way the p-values are 
computed. It seems when I try to do the chi-square test myself, I always 
overestimate the p-value.


The following is a complete example:

silly.dat <- c(2, 1, 5, 3, 6, 4)
dummy.data <- new("aGFF", data = matrix(silly.dat, ncol = 1),
                  annotation = data.frame(Chromosome = 1,
                               Location = c(1, 10, 20, 1000, 1200, 1300)),
                  samples = data.frame(SampleID = 1))

### So we have
cbind(silly.dat, Position = c(1, 10, 20, 1000, 1200, 1300))
     silly.dat Position
[1,]         2        1
[2,]         1       10
[3,]         5       20
[4,]         3     1000
[5,]         6     1200
[6,]         4     1300


do.aGFF.calc(dummy.data, window = 110, thresh = 0.90)
### Cutpoints:
### [1] 5.5
### Thus, all values except 6 (the fifth value) are below the threshold


## Within the window for fifth value we have:
## only value of 6 
do.aGFF.calc(dummy.data, window = 10, thresh = 0.90)@vals[5] ## 0.0877
chisq.test(x = c(0, 1), p = as.vector(table(silly.dat > 5.5)/6)) ## 0.02535

## value of 6 and 4
do.aGFF.calc(dummy.data, window = 202, thresh = 0.90)@vals[5] ## 0.3458
chisq.test(x = c(1, 1), p = as.vector(table(silly.dat > 5.5)/6)) ## 0.2059

## values 6, 4, 3
do.aGFF.calc(dummy.data, window = 402, thresh = 0.90)@vals[5] ## 0.571
chisq.test(x = c(2, 1), p = as.vector(table(silly.dat > 5.5)/6)) ## 0.4386


Where am I computing the chisq in the wrong way?

Thanks,

R.


-- 
Ramón Díaz-Uriarte
Statistical Computing Team
Centro Nacional de Investigaciones Oncológicas (CNIO)
(Spanish National Cancer Center)
Melchor Fernández Almagro, 3
28029 Madrid (Spain)
Fax: +-34-91-224-6972
Phone: +-34-91-224-6900

http://ligarto.org/rdiaz
PGP KeyID: 0xE89B3462
(http://ligarto.org/rdiaz/0xE89B3462.asc)



**NOTA DE CONFIDENCIALIDAD** Este correo electrónico, y ...{{dropped:3}}



More information about the Bioconductor mailing list