data(LahmanData)
# find ID variables in the datasets
IDvars <- lapply(LahmanData[,"file"], function(x) grep('.*ID$', colnames(get(x)), value=TRUE))
names(IDvars) <- LahmanData[,"file"]
str(IDvars)
# vector of unique ID variables
unique(unlist(IDvars))
# which datasets have playerID?
names(which(sapply(IDvars, function(x) "playerID" %in% x)))
################################################
# Visualize relations among datasets via an MDS
################################################
# jaccard distance between two sets; assure positivity
jaccard <- function(A, B) {
max(1 - length(intersect(A,B)) / length(union(A,B)), .00001)
}
distmat <- function(vars, FUN=jaccard) {
nv <- length(vars)
d <- matrix(0, nv, nv, dimnames=list(names(vars), names(vars)))
for(i in 1:nv) {
for (j in 1:nv) {
if (i != j) d[i,j] <- FUN(vars[[i]], vars[[j]])
}
}
d[is.nan(d)] = 0
d
}
# do an MDS on distances
distID <- distmat(IDvars)
config <- cmdscale(distID)
pos=rep(1:4, length=nrow(config))
plot(config[,1], config[,2], xlab = "", ylab = "", asp = 1, axes=FALSE,
main="MDS of ID variable distances of Lahman tables")
abline(h=0, v=0, col="gray80")
text(config[,1], config[,2], rownames(config), cex = 0.75, pos=pos, xpd=NA)
Run the code above in your browser using DataLab