Kaiser-Meyer-Olkin (KMO) Test - How to Interpret Properly

Many sources suggest that a KMO value of larger than .50 suggests the data are appropriate for factor analysis. In this video, I demonstrated that this suggesting is misleading. Instead, on the basis of Kaiser and Rice (1974), I suggest a value of .65 or larger to support the application of a factor analysis.
Kaiser, H. F., & Rice, J. (1974). Little jiffy, mark IV. Educational and psychological measurement, 34(1), 111-117.

[@jessicaturecek9277]

Thank you so much for taking the time to make these videos. They are always a huge help!

[@rbrindamba]

Thank you so much sir.... u proved an information is wealth 😊 keep on helping like this ❤

[@Perfumegal]

Many thanks for this video - it has saved my thesis bacon. xx

[@khanemran2181]

In this video you talk about 0.65 as the minimum KMO value for factor analysis, can you please provide a citation for the same?

[@dsavkay]

Amazing insights! 💯

[@mahamanoumar1802]

Thank you Sir, actually I am using River water samples data. I would like to perform PCA and some other statistical methods but I found KMO coefiicient less than even 0.4. please any help?

[@jiajiantin6140]

Nice! This has been such a big headache for me... every different article I find claims a different number between 0.6 to 0.8

[@PedroRibeiro-zs5go]

Thanks! This video is really nice!

[@nazriesaini4738]

why I run the test but table for KMO and Bartlett test is not appear.. I do click that.. why?

[@roughr4044]

Hey, my p-value for Bartlet's test is 6e-125 which shows it is fit for dimensionality reduction but overall MSA(KMO) = 0.04 which shows is not good for factor analysis. What does this mean?

[@bobbyyankey5967]

Thank you very much for the video Dr. TL Todd. I did not however get the point of 'mixture of positive and negative correlation matrix' right. Again you mentioned for symmetrical data SPSS give a KMO value of 0.5 hence your rejection of value as suggestive of appropriate factor in dataset. My question is is this true for other softwares? Thank you

[@how2stats]

Ha! I'm not Dr. Todd. This is how2stats (Dr. Gignac). I don't think I said symmetrical data. It would be more accurate to say an identity matrix (all .00, except for the variances) yields a KMO value of .50; however, a correlation matrix with a mixture of positive and negative correlations can also yield an average correlation of .00. I don't know about KMO in other programs, but I would be surprised if they calculated it differently.

[@minenhlekhumalo7053]

What if it's 0.515, Is it possible to manipulate data and improve to to 0.65 at least? Just asking.

[@how2stats]

I doubt it. Look at the inter-item correlations, you'll see that they are probably all very low (say, < .09)

[@timiolubiyiph.d.5083]

Many thanks! Informative

[@karlabrionesgaray8111]

what author are you based on to say that it should have a value of .65

[@how2stats]

It's my own recommendation, based on Kaiser's table. You can find some books that recommend .60 or greater, if you Google it.

[@karlabrionesgaray8111]

@@how2statsthanks

[@artint25]

hi, useful video,I am getting chi square value 1.00E4 , can someone help me to figure out the error/problem

[@Abunayeemmdhasan]

May I have the reference link of this paper?

[@how2stats]

Kaiser, H. F., & Rice, J. (1974). Little jiffy, mark IV. Educational and psychological measurement, 34(1), 111-117.

[@sondos2610]

what is the df. value pls ?

[@how2stats]

There are no degrees of freedom to consider, in this context; it's more of an effect size approach.

[@diegoeiras1066]

Great Job! like.

[@AnnaRodriguez-x5s]

Buckridge Parkways

[@WyldeDon-b6u]

0494 Ortiz Fork