This is what I've tried so far and just can't get it. I know I want a value of 3.93 (for Age= y and m) using mahalanobis d as an effect size for a follow up to an MANOVA: age.frame<-data.frame(Age, Friend.Agression, Parent.Agression, Stranger.Agression) > age.frame Age Friend.Agression Parent.Agression Stranger.Agression 1 y 8 7 8 2 y 5 6 8 3 y 6 3 7 4 y 5 5 7 5 m 15 13 10 6 m 13 11 9 7 m 12 12 9 8 m 18 10 7 9 o 11 11 10 10 o 10 4 12 11 o 12 9 12 12 o 9 8 14 13 y 13 7 7 14 y 9 5 10 15 y 11 4 4 16 y 15 3 4 17 m 14 12 8 18 m 10 15 11 19 m 12 11 8 20 m 10 9 9 21 o 10 8 11 22 o 13 11 13 23 o 9 8 12 24 o 7 9 16 ay<-subset(nd,Age=="y") > ay Gender Age Friend.Agression Parent.Agression Stranger.Agression 1 f y 8 7 8 2 f y 5 6 8 3 f y 6 3 7 4 f y 5 5 7 13 m y 13 7 7 14 m y 9 5 10 15 m y 11 4 4 16 m y 15 3 4 > am<-subset(nd,Age=="m") > am Gender Age Friend.Agression Parent.Agression Stranger.Agression 5 f m 15 13 10 6 f m 13 11 9 7 f m 12 12 9 8 f m 18 10 7 17 m m 14 12 8 18 m m 10 15 11 19 m m 12 11 8 20 m m 10 9 9 > ao<-subset(nd,Age=="o") > ao Gender Age Friend.Agression Parent.Agression Stranger.Agression 9 f o 11 11 10 10 f o 10 4 12 11 f o 12 9 12 12 f o 9 8 14 21 m o 10 8 11 22 m o 13 11 13 23 m o 9 8 12 24 m o 7 9 16 > amm<-cbind(am$Friend.Agression, am$Parent.Agression,am$Stranger.Agression) > amm [,1] [,2] [,3] [1,] 15 13 10 [2,] 13 11 9 [3,] 12 12 9 [4,] 18 10 7 [5,] 14 12 8 [6,] 10 15 11 [7,] 12 11 8 [8,] 10 9 9 > aym<-cbind(ay$Friend.Agression, ay$Parent.Agression,ay$Stranger.Agression) > aym [,1] [,2] [,3] [1,] 8 7 8 [2,] 5 6 8 [3,] 6 3 7 [4,] 5 5 7 [5,] 13 7 7 [6,] 9 5 10 [7,] 11 4 4 [8,] 15 3 4 > aom<-cbind(ao$Friend.Agression, ao$Parent.Agression,ao$Stranger.Agression) > aom [,1] [,2] [,3] [1,] 11 11 10 [2,] 10 4 12 [3,] 12 9 12 [4,] 9 8 14 [5,] 10 8 11 [6,] 13 11 13 [7,] 9 8 12 [8,] 7 9 16 > mean(aym) [1] 6.958333 > mean(amm) [1] 11.16667 > mean(aom) [1] 10.375 > ascores<-cbind(Friend.Agression, Parent.Agression, Stranger.Agression) > ascores Friend.Agression Parent.Agression Stranger.Agression [1,] 8 7 8 [2,] 5 6 8 [3,] 6 3 7 [4,] 5 5 7 [5,] 15 13 10 [6,] 13 11 9 [7,] 12 12 9 [8,] 18 10 7 [9,] 11 11 10 [10,] 10 4 12 [11,] 12 9 12 [12,] 9 8 14 [13,] 13 7 7 [14,] 9 5 10 [15,] 11 4 4 [16,] 15 3 4 [17,] 14 12 8 [18,] 10 15 11 [19,] 12 11 8 [20,] 10 9 9 [21,] 10 8 11 [22,] 13 11 13 [23,] 9 8 12 [24,] 7 9 16 > mean(ascores) [1] 9.5 > S<-cov(ascores) > S Friend.Agression Parent.Agression Stranger.Agression Friend.Agression 10.476449 4.461957 -2.003623 Parent.Agression 4.461957 10.940217 3.489130 Stranger.Agression -2.003623 3.489130 8.427536 > ymean<-mean(aym) > mmean<-mean(amm) > omean<-(aom) > mahalanobis(aym,amm,S) [1] 8.223093 18.633617 9.838251 6.301238 7.958673 4.707784 2.547173 6.671211 Warning message: In sweep(x, 2, center) : STATS is longer than the extent of 'dim(x)[MARGIN]'
From: tyler_rin...@hotmail.com To: r-help@r-project.org Date: Mon, 21 Mar 2011 20:33:58 -0400 Subject: [R] Using the mahalanobis( ) function Hello all, I am a 2 month newbie to R and am stumped. I have a data set that I've run multivariate stats on using the manova function (I included the data set). Now it comes time for a table of effect sizes with significance. The univariate tests are easy. Where I run into trouble filling in the table of effect sizes is the Mahalanobis D as an effect size. I've included the table so you can see what group's I'm comparing. I know there's a great function for filling in ?1 and ?2 : mahalanobis(x, center, cov, inverted=FALSE, ...) I need to turn the sub groups scores for y (young), m (middle) and o (old) into clusters. The problem is I lack the knowledge around cluster analysis of what goes into the function for [x, center, & cov.] I have only a basic understanding of this topic (a picture of a measured distance between two clusters on a graph). I think I have to turn the data into a matrix but lack direction. Could someone please use my data set or a similar one (a multivariate with at least 3 outcome variables) and actually run this function (mahalanobis). Then please send me your output from [R] starting from the data set all the way to the statistic. PS I know the Mahalanobis D should be ?1=3.93 & ?2=3.04. Ive read and reread the manual around mahalanobis() and have searched through the list serve for information. The info is for people who already have a grasp of how to implement this concept. I am running the latest version of R on a windows 7 machine. Effect Sizes Contrasts| Dependent Variables Friends Parents Strangers All Young-Middle -1.8768797* -3.2842941* -1.1094004* ?1 Middle-Old 1.34900725* 1.54919532* -2.0107882* ?2 (sorry the column names and values dont line up) Age Friend Agression Parent Agression Stranger Agression y 8 7 8 y 5 6 8 y 6 3 7 y 5 5 7 m 15 13 10 m 13 11 9 m 12 12 9 m 18 10 7 o 11 11 10 o 10 4 12 o 12 9 12 o 9 8 14 y 13 7 7 y 9 5 10 y 11 4 4 y 15 3 4 m 14 12 8 m 10 15 11 m 12 11 8 m 10 9 9 o 10 8 11 o 13 11 13 o 9 8 12 o 7 9 16 [[alternative HTML version deleted]] ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. [[alternative HTML version deleted]]
______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.