Rather than simply plotting the distribution, let's plot a histogram of the distribution so that we can see what the approximate error bins will look like. In fact, let's generate two different histograms, one containing the actual counts and the other containing percentages:

par(mfrow=c(1,2)) h <- hist(x,freq=FALSE,right=FALSE,breaks=c(-.20, -.15,-.10,-.05, 0, .05, .1, .15, .20)) print(h) plot(h$mids,100*h$counts/sum(h$counts),type="h",lwd=25, lty=1, main="Probability of Errors Introduced")

In the plots below, you can see from the histogram on the left that there are no values beyond the limits specified. In fact, since the distribution was generated with a mean=0, a good number of the test elements ...