equal to 1 or not equal to 1, that's the question....

[Oleg Moskvin]

I just cannot avoid sharing this crazy thing...

I've derived a "p-value-like" statistics in a slightly weired way (that's a separate story) to feed it into Bioconductor's "piano" package.

> str(statistics)
'data.frame':    3911 obs. of  1 variable:
 $ pNG.2v1: num  0.852 0.512 0.865 0.798 0.647 ...

> summary(statistics)
    pNG.2v1     
 Min.   :0.0000 
 1st Qu.:0.4859 
 Median :0.8501 
 Mean   :0.7053 
 3rd Qu.:0.9963 
 Max.   :1.0000 

# Bioconductors "piano" package doesn't like this as a p-value equivalent:
### Checking arguments...Error in checkLoadArg(geneLevelStats, directions, geneSetStat, signifMethod,  :
###  geneLevelStats does not lie in [0,1], geneLevelStats can only be p-values for geneSetStat='fisher'

# if we look at the checkLoadArg function from piano package
# http://search.bioconductor.jp/codes/11371
# the trouble is here:

> min(statistics) >= 0 & max(statistics) <= 1
[1] FALSE

# what???

# this was narrowed down further... and now I am speechless:

> max(statistics)
[1] 1

> max(statistics) <= 1
[1] FALSE

> max(statistics) == 1
[1] FALSE


# when the shock was over, I started to suspect what is going on:

> round(max(statistics),0) == 1
[1] TRUE

> max(statistics) > 1
[1] TRUE

> max(statistics) - 0.000000000000001 > 1
[1] FALSE

> max(statistics) - 0.0000000000000001 > 1
[1] TRUE

# Bingo!