[Carlo Lazzaro]

Sandeep;
I share both your points with the proviso that VIF is considered a bit oversold.

[sandeep kaur]

Carlo Lazzaro
Thanks for explaining it

Crux is:
1) Do the correlation analysis.

2) If high correlation coefficients: keep or drop the variables as per your modeling/other tests requirements. For examples it two predictors are correlated,drop one variable. if predictor and outcome is correlated suggests there is a strong association.

3) Higher the correlation coeffiecient/correlation matrix of predictors: It indicates there are chances of colinearity, but does not imply there will be colinearity. It needs further investigation

4) Presence of multicolinearity:
-VIF is only one way but it's oversold
- Confirm with other tests if required
- STATA drops the multicolinear variables automatically from modeling


P.S.
Appreciate your help throughout-sorry for bugging with these small questions. I wanted to clear concepts, it can be confusing with so much information floating around.

Regards
Sandeep

[Carlo Lazzaro]

Sandeep:
1) Ok;
2) If two predictors are quasi-perfectly multicollinear you do not have to drop one of them necessarily. It depends, among othre things, on the multicollinearity threshold value you consider as critical (0.7 might be of some concerns) and on the bearing on standard errors (and related stuff);
3) Ok;
4) -estat vce, corr- post -regress-like commands.

[sandeep kaur]

Carlo Lazzaro >Thanks for clarifying it.

[sandeep kaur]

Nick Cox

Collinearity:
a) I have mix of both binary and categorical independent variables. There are no continuous variables.

Pearson can be used to check correlation when both variables are binary.
Can it be used a) if both variables are categorical b) if one variable is binary and second variable is categorical?

b) How about using pairwise correlation coefficient matrix in a scenario where there is a mix of binary and categorical independent variables?

Multicollinearity:
a) Can vif be used to check for multicollinearity when there is a mix of binary and categorical independent variables?

[Nick Cox]

sandeep kaur You seem to be asking the same basic questions repeatedly with minor variations. I'd consult your teachers on the best books for you to read or the best course for you to take.

[sandeep kaur]

Nick Cox Thanks for your help.