Collin and VIF

[Zuraida Rizki]

Dear statalist, i just tried to find the collinearity between the regressor using "collin" and "vif, uncentered" command after xtreg. the vif value is different between the two. which one is the true vif? Thanks in advance

[Carlo Lazzaro]

Zuraida:
to jump that issue altogether, you may want to take a look at: https://www.statalist.org/forums/for...ticollinearity

[Zuraida Rizki]

Hey Carlo, thanks for replying

i've read this source from your suggestion link https://www3.nd.edu/~rwilliam/stats2/l11.pdf , in that writing they just give the example of 2 independent variable, and the result from "vif" and "collin" did not differ. but in my case, if i run "vif uncentered" after xtivreg, the vif value is super high, meanwhile if i try the "collin" command, the vif value change to super small values. which one should i trust? can you help me?

Here is, my result

. xtivreg lnPDRBkap lnPMTB GE1 TPT (AHH RLS=Kes1 Pend1 Miskin CRT1), re ec2sls small

EC2SLS random-effects IV regression             Number of obs      =       231
Group variable: provnum                         Number of groups   =        33

R-sq:  within  = 0.8111                         Obs per group: min =         7
       between = 0.3090                                        avg =       7.0
       overall = 0.3202                                        max =         7

                                                F(5,226)           =    109.97
corr(u_i, X)       = 0                          Prob > F           =    0.0000

------------------------------------------------------------------------------
   lnPDRBkap |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         AHH |   .0693889   .0439696     1.58   0.116    -.0172538    .1560316
         RLS |   .2334762   .0873845     2.67   0.008     .0612837    .4056688
      lnPMTB |   .0173029   .0088865     1.95   0.053    -.0002081     .034814
         GE1 |   .0031024   .0012392     2.50   0.013     .0006606    .0055442
         TPT |  -.0087449   .0033184    -2.64   0.009    -.0152838    -.002206
       _cons |   3.328132   2.393156     1.39   0.166     -1.38762    8.043884
-------------+----------------------------------------------------------------
     sigma_u |  .64496156
     sigma_e |  .06747021
         rho |  .98917496   (fraction of variance due to u_i)
------------------------------------------------------------------------------
Instrumented:   AHH RLS
Instruments:    lnPMTB GE1 TPT Kes1 Pend1 Miskin CRT1
------------------------------------------------------------------------------

. vif, uncentered

    Variable |       VIF       1/VIF  
-------------+----------------------
         AHH |    605.17    0.001652
      lnPMTB |    468.83    0.002133
         RLS |    104.31    0.009587
         TPT |      9.53    0.104884
         GE1 |      2.61    0.382566
-------------+----------------------
    Mean VIF |    238.09

. collin AHH RLS lnPMTB GE1 TPT
(obs=231)

  Collinearity Diagnostics

                        SQRT                   R-
  Variable      VIF     VIF    Tolerance    Squared
----------------------------------------------------
       AHH      1.84    1.35    0.5447      0.4553
       RLS      1.62    1.27    0.6185      0.3815
    lnPMTB      2.51    1.59    0.3979      0.6021
       GE1      1.98    1.41    0.5053      0.4947
       TPT      1.33    1.15    0.7513      0.2487
----------------------------------------------------
  Mean VIF      1.86

[Zuraida Rizki]

anyone who can help me?

[Richard Williams]

Why are you using uncentered? What happens if you don't use that option?

In any event I am not sure that collin and vif are going to produce identical results when you are using xtivreg.

[Zuraida Rizki]

Hey Richard thanks for replying, because if i just run "vif" there was an error message and i think the "vif" is not suitable after xtivreg. If i run "vif" the error massage is "not appropriate after regress, nocons;
use option uncentered to get uncentered VIFs"

any idea what am i supposed to use for checking the multicollinearity? because i really have to did that for my final assignment. Thank you in advance Richard!

[Zuraida Rizki]

anyone who can help me?