SPSS Essentials: Mahalanobis Distance Test in SPSS for Finding Multivariate Outliers

When you are cleaning your raw data, you will want to check for outliers. The Mahalanobis Distance test identifies multivariate outliers. This video will teach you the steps to doing a Mahalanobis test in SPSS, how to compute the probabilities for the Mahalanobis scores, sort appropriately, decide which cases to exclude from further analysis, and write up the results in APA style.
This video is an abbreviated version of a longer video that also includes a review of outliers, an explanation of multivariate outliers, the use of the Chi-square distribution table, choosing cutoff scores for exploratory data analysis, and a fuller explanation of the theory behind Mahalanobis.
Full video: • Finding Multivariate O...
This video teaches the following commands and techniques:
Mahalanobis test
Compute Variable
Tabachnick, B. G., & Fidell, L. S. (2013). Using multivariate statistics (6th ed.). Boston, MA: Allyn and Bacon.
SYNTAX:
COMPUTE pMAH_1= 1 - CDF.CHISQ(mah_1,8).
EXECUTE.
SORT CASES BY pMAH_1 (A).
This series uses the data set Mahalanobis_PH.sav
Link to a Google Drive folder with all of the files that I use in the videos including the dataset.
drive.google.c...

[@naylianas8159]

Good quality video n good explanation..

[@lilgummybear11]

Thank you, this helped so much

[@ResearchByDesign]

You're welcome! Glad that it helped

[@syedabbas7720]

You didn't explain how to extract the table of the results from SPSS to the thesis

[@cecyliaadamczak4301]

Hi Professor Dan, Just to clarify, do you include the outcome variable measured by the surveys in the MD analysis?

[@ResearchByDesign]

Yes...if you have four predictors and one outcome, then your "degrees of freedom" for the Mahalobis would be 5 (not n-1). Best!

[@zubairawan9088]

Why have you selected the p-value to be 0.001? Please explain.

[@hayatiibrahim6822]

hi prof, do we need to compute the variables first before conducting the Mahalanobis?

[@ResearchByDesign]

Yes, you do. Combine the sub scale items into a single subscale and that is one variable. Do the same for other sub scales and your DV. The df for Mahalanobis is the total number of variables (not n - 1) Good luck

[@hameemislam753]

Sir my moderating variable is gender i.e categorical should i also take that for checking outliers

[@ResearchByDesign]

Hello Hameem - you will not need to check for outliers on the Gender variable (assuming it is dichotomous), but you should consider evaluating your DV both as the entire variable, then split by Gender. Good luck!

[@a.w.n_tn]

thank you for sharing ❤

[@ResearchByDesign]

You are so welcome

[@pamand1973]

Dear Prof Dan, Are the mediating variable and moderating variables included in the Mahalanobis test

[@ResearchByDesign]

Yes, every IV and DV variable (make sure they are aggregated into total scores, not individual items) are included. The df is the total number of variables, not n- 1. Good luck

[@caglacobanoglu3609]

@@ResearchByDesign while you mean variables, you also include your dvs right?

[@bandungmee]

@@ResearchByDesign Prof, I have 7 IVs and 1 mediator - that means while doing MD analysis, my total IVs considered as 8. is it correct?

[@mikorees5853]

Dear Professor Dan, I am running Mahalanobis distance analysis but some cells are appearing empty in Mahalanobis column. Kindly suggest what should be done when Mahalanobis distance cell is empty?

[@ResearchByDesign]

The empty cells are due to missing data in the variables you are using in the Mahalanobis test. If a person has a missing score on any variable (in your model - the ones you chose for the test), you will not get a Mahalanobis score for that individual.

[@mikorees5853]

@@ResearchByDesign thank you Prof Dan

[@bandungmee]

@@ResearchByDesign Dear Prof, I am also getting the same null value after MD analysis. However, this was done before missing value imputation. Is this the correct way? Or do I need to perform imputation and run the MD analysis again? thank you

[@fn5k]

Cool shirt

[@ResearchByDesign]

Thanks! I wore the shirt for a series of videos a few years ago and it kind of became a thing.

[@syedabbas7720]

You didn't explain how to extract the table of the results from SPSS to the thesis