Understanding and Identifying Multicollinearity in Regression using SPSS

This video explains multicollinearity and demonstrates how to identify multicollinearity among predictor variables in a regression using SPSS. Correlation, tolerance, and variance inflation factor (VIF) are reviewed.

[@GChan129]

I have been watching some of your videos about narcissism recently. Independently I needed to research videos for my statistics exam, came across this video and thought "that voice is familiar..."
Thank you for your help in understanding narcissists as well as statistics!

[@MHlifestyle_60]

Sir, your videos are very helpful, i appreciate your effort.

[@debbiekane7977]

Thanks you! Very easy to understand the way you explained it.

[@kosalaudugahapattuwa1909]

Very useful to understand the Multimillionearity in Regression and Thanks.

[@douglaswilliams5794]

These helped me get my head around the multiple regression analysis I am doing in my dissertation. Thanks for posting these.

[@thawornlorga1972]

Great values. Thank you very much for the video

[@dubegemeda7614]

Hi Dr. Thank u very. I have got alot of information from your video.

[@risausa4796]

Thanks for this helpful video!

[@sanera2008]

Thank you and this is very helpful!

[@DrGrande]

You are quite welcome!

[@carmenpower2370]

Thank you so much for making it so simple to understand

[@DrGrande]

You're welcome - thank you for watching -

[@dimuthpeiris1322]

Thank you very much, Dr Grande. In logistic regression, if many of the input variables are either yes or no is it necessary to run colinearity assessment before running the regression analysis

[@aymanzein7]

Mumtaz ! = Excellent !

[@aycaakdilsonmez1036]

Hi. Thank you for the explanation. What if we are dealing with latent variables, such as perceived image, which contains 5 separate observed variables and customer trust, which contains 6 separate observed variables? Shall we treat latent variables the same as observed variables in order to solve collinearity problem?

[@intandianna4825]

THANK YOU DR. GRANDE.

[@DrGrande]

You're welcome -

[@krijgsmand77]

Super usefull! Thanks from the Netherlands

[@DrGrande]

You're welcome!

[@selamawitguta4756]

Thank you very Much!!!!!!

[@lindenmueller2824]

Hahaha.... "Now the depression and hopelessness variables..."
Your choice of variables lends some dry humor to what is also a helpful tutorial. Thank you!

[@DrGrande]

You're welcome -

[@jameslambert6596]

Firstly: This video is really helpful - thank you!
What should I do if my eigenvalues (from the Collinearity Diagnostics table) disagree with the tolerance and VIF? I have four predictors, each with VIF < 2 and tolerance < .66. This would suggest no multicollinearity. However, a couple of the eigenvalues are very close to 0, which would suggest multicollinearity.

[@sabbirsakib8483]

Thank you sir.

[@Ashtangi83]

Hi Dr. Grande, thank you so much for this awesome video and explanation. I'd like to reference you, but youtube isn't the most reliable scientific source. Have you published this information anywhere? Thank you!

[@baoair]

Thanks a lot for this great video

[@DrGrande]

You're welcome - thanks for watching -

[@elizabethnundu1865]

thank you for simplifying it

[@DrGrande]

You're welcome and thank you for watching -

[@leilaamini405]

awesome thanks

[@emantoraih7700]

Thank you

[@jirir7527]

Thanks for your videos man. But I do have a question. I have multicolli but I dont want to leave the variable out of it, I want to correct this. Is this possible by using a dummy for this variable?

[@abebemenberu7126]

Dear Dr. Grande, I have data for a model comprising of multiple IV, mediating variables and multiple dependent variables. How do I compute multicollinearity? Thank you very much.

[@oracklemaeene7334]

Thank you very much for this helpful video. Given the different cut-offs for VIF (2, 3 or 10) that are used, have you got a source of information that states which cut-off would be most appropriate? Also, are eigenvalues useful in detecting multicollinearity? If yes, how should they be interpreted?

[@oracklemaeene7334]

Also, can VIF, tolerance and eigenvalues be used for independent variables which are not normally distributed?

[@nadimbader6144]

Hi
First, thank you for explaining multicollinearity
In my case, i have 8 independent variables and here is the Coefficients table.
Model Collinearity Statistics
Tolerance VIF
var1 .186 5.374
var2 .487 2.055
var3 .325 3.081
var4 .150 6.679
var5 .344 2.911
var6 .358 2.790
var7 .542 1.844
var8 .707 1.414
Based on this table, I removed var4 from the Linear Regression model.
Before removing it, the R-square value was 0.277, but after removing var4 the R-Square value become 0.266.
Is that ok? should I keep var4 variable or not?
Can you please explain this?
Regards,
Nadeem Bader

[@talkiteasy7219]

Thank youuuu doctor!! But please what does it mean when two independent variables have the same VIF and tolerance!? I will be grateful if you answer me! It's kinda urgent

[@zulffiquer732]

Thanks

[@DrGrande]

You're welcome, thanks for watching -

[@afrinews6949]

helpful

[@avinkothari]

Sir, could you give us the link to data so we can practise it?

[@akashanee3048]

Hello Dr. Grande, thank you for the video!!! it was realy useful for me. One question remains: i have different types of variables in my binominal logistic regression.
Am i allowed to use the VIF for all types of predictor variables, even if they are mixed (dichotom (1/0), ratio like Age (0-50), ordinal (1=low, 2=middle, 3=high)? My dependent Variable is also dichotom (0/1).
is it allowed to run a spearman correlation to check the correlations between all these different types of variables? Pearson is not allowed, because of the mixed vales. Which correlation courld i run in SPSS to see the correlation between thes different variabletypes? Because the VIF just tells me that there is a mulitcollinarity but not between which variabales...so without a correlation first i will have no clue between which variables the multicollinarity exists, as far as i understood.
Thank you very much!!! Kind regrads Akashanee

[@zulffiquer732]

I think in your situation, you need to use Logistic Regression Model. First, you keep all your IVs in IV Box. One by one, shift each IV from the IV box to DV box, check Multicoliearity from statistics option, click OK.

[@quitatate]

Dr. Grande,
Can you solve multicollinearity issues using the stepwise regression method?

[@forrestwall9858]

Laquita Tate
Did you every find the answer to your question?

[@MrSupernova111]

I'm not an expert but my guess is no because I ran a stepwise regression in SAS and I got the same output I did in Excel. As of right now, my solution is to delete the variables with highest VIF values and play with the data a bit.

[@forrestwall9858]

MrSupernova111
I agree with you. 1. Set up a scatter plot to identify Multicollineary, 2. Look at the standard regression table to spot Muticollineary, 3. Look at the regression sheet and then coefficients 4. Pull one of the independent variables out of the samples and rerun your regression.

[@saikiraraju.m.r]

what about categorical and continuous variables?

[@libertarianPinoy]

You didnt explain what tolerance was.