Convert an object of class "dist" to a data frame in r

Question

if it is possible to convert a data frame to an object of class "dist", can it be done just the opposite? convert class "dist" to data frame? eg

<distance (Hasil)

1 2 3 4 2 0.088814413 3 0.084929382 0.030413813 4 0.063245553 0.029120440 0.044418465 5 0.061983869 0.027018512 0.036400549 0.009055385 

and the result in a data frame

<

  col row distance 1 2 0.088814413 1 3 0.084929382 1 4 0.063245553 1 5 0.061983869 2 3 0.030413813 2 4 0.029120440 2 5 0.027018512 3 4 0.044418465 3 5 0.036400549 4 5 0.009055385 

Answer 1

 library(maps) data(us.cities) d <- dist(head(us.cities[c("lat", "long")])) ## 1 2 3 4 5 ## 2 20.160489 ## 3 23.139853 40.874243 ## 4 15.584303 9.865374 38.579820 ## 5 27.880674 7.882037 48.707100 15.189882 ## 6 26.331187 41.720457 6.900101 41.036931 49.328558 library(reshape2) df <- melt(as.matrix(d), varnames = c("row", "col")) df[df$row > df$col,] ## row col value ## 2 2 1 20.160489 ## 3 3 1 23.139853 ## 4 4 1 15.584303 ## 5 5 1 27.880674 ## 6 6 1 26.331187 ## 9 3 2 40.874243 ## 10 4 2 9.865374 ## 11 5 2 7.882037 ## 12 6 2 41.720457 ## 16 4 3 38.579820 ## 17 5 3 48.707100 ## 18 6 3 6.900101 ## 23 5 4 15.189882 ## 24 6 4 41.036931 ## 30 6 5 49.328558 

Answer 2

I would really write a function something like this:

 myFun <- function(inDist) { if (class(inDist) != "dist") stop("wrong input type") A <- attr(inDist, "Size") B <- if (is.null(attr(inDist, "Labels"))) sequence(A) else attr(inDist, "Labels") if (isTRUE(attr(inDist, "Diag"))) attr(inDist, "Diag") <- FALSE if (isTRUE(attr(inDist, "Upper"))) attr(inDist, "Upper") <- FALSE data.frame( row = B[unlist(lapply(sequence(A)[-1], function(x) x:A))], col = rep(B[-length(B)], (length(B)-1):1), value = as.vector(inDist)) } 

Now imagine what we start with (note the non-numeric row and column names):

 dd <- as.dist((1 - cor(USJudgeRatings)[1:5, 1:5])/2) # CONT INTG DMNR DILG # INTG 0.56659545 # DMNR 0.57684427 0.01769236 # DILG 0.49380400 0.06424445 0.08157452 # CFMG 0.43154385 0.09295712 0.09332092 0.02060062 

We can change it simple:

 myFun(dd) # row col value # 1 INTG CONT 0.56659545 # 2 DMNR CONT 0.57684427 # 3 DILG CONT 0.49380400 # 4 CFMG CONT 0.43154385 # 5 DMNR INTG 0.01769236 # 6 DILG INTG 0.06424445 # 7 CFMG INTG 0.09295712 # 8 DILG DMNR 0.08157452 # 9 CFMG DMNR 0.09332092 # 10 CFMG DILG 0.02060062 

---

Quick performance comparison:

 set.seed(1) x <- matrix(rnorm(1000*1000), nrow = 1000) dd <- dist(x) ## Jake function fun2 <- function(inDist) { df <- melt(as.matrix(inDist), varnames = c("row", "col")) df[as.numeric(df$row) > as.numeric(df$col), ] } all(fun2(dd) == myFun(dd)) # [1] TRUE system.time(fun2(dd)) # user system elapsed # 0.346 0.002 0.349 system.time(myFun(dd)) # user system elapsed # 0.012 0.000 0.015 

Answer 3

I would do something like this.

 library(reshape) set.seed(123) mat <- matrix(rnorm(25),ncol=5,nrow=5,byrow=T) # Creating a 5 X 5 matrix d <- dist(mat) df <- melt(as.matrix(d)) #Converting the dist object to matrix while using melt p <- t(apply(df[,c(1,2)],1,FUN=sort)) rmv1 <- which(p[,1] == p[,2]) p <- paste(p[,1],p[,2],sep="|") rmv2 <- which(duplicated(p)) df <- df[-c(rmv1,rmv2),] #removing self distances and duplicated distances df X1 X2 value 2 1 3.812287 3 1 2.311832 4 1 4.385048 5 1 2.854179 3 2 1.916895 4 2 1.557744 5 2 2.880357 4 3 2.509214 5 3 2.886526 5 4 3.408147