Linear regression - Detecting multicollinearity

[Tom Hsiung]

I searched the Internet and found the syntax of 'estat vif'. But there are three methods to do this job and I don't know how via Stata.

My output

. estat vif

    Variable |       VIF       1/VIF  
-------------+----------------------
1.ttestfor~s |      1.12    0.889984
1.agefortt~t |      1.85    0.541637
1.gendercode |      1.23    0.811343
        1.af |      1.56    0.639049
1.hyperten~n |      1.17    0.858267
1.non11orn~1 |      1.08    0.925788
1.nonaaorn~a |      1.05    0.949364
       1.chf |      1.14    0.879495
1.hypoalbu~t |      1.11    0.901499
1.akifortt~t |      1.18    0.845265
1.poten~eddi |      1.13    0.885358
1.ttestind~n |      2.76    0.362166
1.ttesttar~r |      2.21    0.452392
1.afandwar~y |      1.27    0.786499
-------------+----------------------
    Mean VIF |      1.42

.

[Carlo Lazzaro]

Tom:
authoritative contributors to this forum consider (quasi-extreme) multicollinearity as a bit oversold stuff.
At least one pretty famous econometric textbook (https://www.amazon.com/Course-Econom.../dp/0674175441) devote some pages to the (ir)releveance of that issue.
That said, possibly Paul Allison (https://uk.sagepub.com/en-gb/eur/mul...ssion/book8989) delves into alternative approach to test for multicollinearity.
In my opinion, I think that -estat vif- does exactly what expected and I do not see the need to some other commands.
Anyway, you can type -seacrh vif- and see what Stata gives you back.

[Richard Williams]

If you want more, I like Phil Ender's collin, available from UCLA.

net describe collin, from(https://stats.idre.ucla.edu/stat/stata/ado/analysis)

You can also see Allison's discussion at

https://statisticalhorizons.com/multicollinearity

[Tom Hsiung]

Got it! Thanks. I like this forum more and more!

Tom

[Marcos Almeida]

If you like this forum, please follow the advice on not to share snapshots of material with potential copyright concerns. In case the material is free to use, please inform the source as well as the author, accordingly. Thanks.