[BioC] odd results from vennCounts()

[Gordon Smyth]

At 07:33 PM 2/09/2004, Jacqueline Hall wrote:
>Hello everyone,
>
>I have been getting some stange results from vennCounts() package limma,
>without any assocaited warnings.
>
>example:
>
>try <- matrix(9,nrow=10,ncol=3)
>try[10,3] <- 0.2
>try[1,2] <- 0.2
>m <- classifyTestsF(try)
>a <- vennCounts(m)
># vennDiagram(a)

What do you think is strange? When I run it, the venn diagram seems a 
perfectly sensible reflection of the data you gave it. You gave it all 
large numbers (9 is pretty large for a std normal) except for two values, 
so it classified all the large values as significant and the two low values 
as non-significant. What else could you expect it to do?

Gordon

>If i've understood correctly,it seems that this occurs when one of the
>groups is all 1's / all "up".
>If you add in try[2,1] <- 0.2 to the above code ie. break the homogeneity of
>that "all 1" group, it seems to give sensible results again.
>Am i making some kind of mistake to get these odd results or should the
>function be printing some kind of warning about a homogenous group, so that
>i know not to trust the results?
>
>Thanks for your help,
>
>Jacqueline