Multivariate Outlier Detection

[monhaml]

Hi all,

I want to compare the results of "classical multivariate mahalanobis distance" scores of 5 variables for 2 million cases and "minimum covariance determinant based mahalanobis distance" scores at 0.001 and 0.01 level respectively. Also I want to see this difference at the plot. In addition I want the outlier numbers and the cases which are outliers as a table.

When I searched the R documents regarding this issue, I saw some packages with manuals but I think they did not give an step by step example. So it makes the procedure difficult for me. Do you have suggestions or codes?

 

[bugman]

Have you read the documetation for mvoutlier?

There are examples in there.

http://cran.r-project.org/web/packages/mvoutlier/mvoutlier.pdf

 

[monhaml]

Yeah I have taken a look. But some issues are missing:

1- While plotting chi square, I want it 0.001 level and 0.01 both at same plot.

2- Also I want to see outlier data table. How can I call it? For instance, I am working on only 5 variables but my data set includes 7 variables. I need outliers table with 7 variables..