In R, fisher.test returns different results if I use vectors vs contingency table

Question

Using R, I noticed that I get different results if I use fisher.test() with raw data (as factors) versus a contingency table. The documentation states that fisher.test(x,y) will compute a contingency table from x, y by treating them as factors.

My example is a baseline statistics exercise: comparing the gender split in a placebo group to the experimental group.

Here's my example for running fisher.test on factors:

> expData <- data.frame(Placebo=factor(c("Y","Y","Y","Y","N","N","N","N")),
    Gender=factor(c("F","F","F","M","M","M","M","F")))
> expData
  Placebo Gender
1       Y      F
2       Y      F
3       Y      F
4       Y      M
5       N      M
6       N      M
7       N      M
8       N      F
> fisher.test(expData[expData$Placebo=="Y","Gender"],
        expData[expData$Placebo=="N","Gender"])

        Fisher's Exact Test for Count Data

data:  expData[expData$Placebo == "Y", "Gender"] and 
        expData[expData$Placebo == "N", "Gender"] 
p-value = 0.25
alternative hypothesis: true odds ratio is not equal to 1 
95 percent confidence interval:
  0.00000 13.00002 
sample estimates:
odds ratio 
         0 

Here's my example for building a contingency table representing the above data and running fisher.test on that:

> contingency <- matrix(c(3,1,1,3),nrow=2,
     dimnames = list(tx=c("Placebo","Exper"), gender=c("F","M")))
> contingency
         gender
tx        F M
  Placebo 3 1
  Exper   1 3

> fisher.test(contingency)

        Fisher's Exact Test for Count Data

data:  contingency 
p-value = 0.4857
alternative hypothesis: true odds ratio is not equal to 1 
95 percent confidence interval:
   0.2117329 621.9337505 
sample estimates:
odds ratio 
  6.408309 

Am I using fisher.test in the wrong way? Or maybe building my contingency table wrong?

Answer 1

The first test should read

> fisher.test(expData[,1], expData[,2])

    Fisher's Exact Test for Count Data

data:  expData[, 1] and expData[, 2] 
p-value = 0.4857
alternative hypothesis: true odds ratio is not equal to 1 
95 percent confidence interval:
 0.001607888 4.722931239 
sample estimates:
odds ratio 
  0.156047 

as per the doc: x is the outcome and y is the factor (or vice-versa).