Thread: conversion mahalanobis distance (multivariate distributions) to percentage

1. Originally Posted by aznimah
and i have coordinate data of vertex points. here is my data link image : FreeImageHosting.net Hosting Service
I don't understand how you can get distances from this.

For each group you need a separate mean vector and covariance matrix. I'm sorry we seem to be going in cirlces here.

2. Originally Posted by pickslides
I don't understand how you can get distances from this.

For each group you need a separate mean vector and covariance matrix. I'm sorry we seem to be going in cirlces here.
yes, its ok since you willing to spent your time answering my questions. really appreciate that.

i clarify it more details:
i have 20 dataset of Diamond Shape. Since the Diamond shape has 4 vertex points or corner points, therefore it will be 4 coordinates (x,y), you can see in my image attachment link here :http://www.freeimagehosting.net/imag...ee8068dc6a.png

my comparison using mahalanobis distance would be between Group A (reference image coordinate) and Group B (user image coordinate)
each group have same size of variable=refering to numbers of column (x,y)
but having a different number of row (vertex coordinate) since i applied masking on reference image coordinate
so make the data
36 coordinate (x,y) Group A (referEnce image)
4 coordinate (x,y) Group A (User image)

then i calculate the mahalanobis distance between 2 groups by refering on this site : Mahalanobis Distance

the overall, i will have 20 dataset which have 20 results of mahalanobis distance as i place on thik image : http://www.freeimagehosting.net/imag...a01a1d65e8.png
please ignore the coordinate (x1,y1) and (x2,y2) on the image, just i want to show the mahalanobis distance for every diamond shape.

what i need to do is, how can i measure in term of percentage that each result of mahalanobis distance that i get earlier into example 69%, 80%?
or there is another way that i can interprate the mahalanobis distance value to be conclude?