qBCI.RdBins numeric variables according to their quantiles and computes the Bertin Classification Index BCI.
The data.frame method computes the multivariate qBCI and not the pairwise values (c.f. cmat).
qBCI(x, ...) # S3 method for default qBCI(x, y, p = NULL, k = 5, iter=20, …) # S3 method for data.frame qBCI(x,p = NULL, k = 5, sort = TRUE, iter=20, …)
| x | A numeric vector (in this case |
|---|---|
| y | A numeric vector. |
| p | A percentage to use for the quantiles sequence. See details. |
| k | A minimum expected number of observations in each cell after the binning. |
| sort | Whether or not to compute the BCI for the optimized tables or not. If not, kendalls is usually a better alternative. |
| iter | An optile parameter. |
| … | dots |
The breakpoints for the binning are the data quantiles according to equidistant probabilities seq(0,1,p) where p is minimal under the condition that each cell has an expected number of observations of at least k.
A value between 0 and 1.
# NOT RUN { qBCI(rnorm(100),runif(100)) # non-functional relationship: x1 <- runif(500,0,10) x2 <- runif(500,0,10) y1 <- x1+rnorm(500,sd=1) y2 <- 10-x2+rnorm(500,sd=1) x <- c(x1,x2) y <- c(y1,y2) plot(x,y, pch = 19) wdcor(x,y) 1 - qBCI(x,y) y1 <- x1+rnorm(500,sd=0.1) y2 <- 10-x2+rnorm(500,sd=0.1) x <- c(x1,x2) y <- c(y1,y2) plot(x,y, pch = 19) wdcor(x,y) 1 - qBCI(x,y) # or a quadratic curve: test <- sapply(seq(0,4,0.2),function(s){ x <- runif(200,-1,1) y <- 5+12*x^2+rnorm(200,sd=s) return(c(cor(x,y), wdcor(x,y), 1 - qBCI(x,y))) }) plot(test[3,],type="l", ylim=c(-0.2,1)) lines(test[1,], col = 2, lwd = 2) lines(test[2,], col = 3, lwd = 2) # }