248236_en-US

liliconcom

Differences in Mahalanobis results -- Multivariate vs. Screening platforms

I've been reading up on Mahalanobis distances, and discovered that there are 2 places to calculate Mahalanobis in JMP:

 

• Analyze/Multivariate Methods/Multivariate, then from dropdown select Outlier Analysis/Mahalanobis Distances.
• Analyze/Screening/Explore Outliers, then choose Multivariate Robust Outliers.

These two methods give significantly different results for the same data inputs.

 

Anybody know why?  I'm attaching a sample file with scripts for both options.

Re: Differences in Mahalanobis results -- Multivariate vs. Screening platforms

Multivariate Robust Outliers in the Explore Outliers platform calculates the Mahalanobis distances from the estimated robust center and variance. To match them to the Multivariate platform's Outlier analysis, select the Robust estimation method in the launch window.

Re: Differences in Mahalanobis results -- Multivariate vs. Screening platforms

Thank you Jian!  This wasn't obvious to me.

 

I was able to calculate the Multivariate Mahalanobis distance, but couldn't do the Outlier Analysis Mahalanobis.  After searching for help on the Robust option in the Multivariate Launch Window, I can now see the difference in the calculations.

 

If anyone else is interested, here is a link to the Robust Help.