0
$\begingroup$

enter image description here

As you can see there vif for all variables is too high...My interpretation is that all variables are highly correlated to each other. How can I further investigate on this? Or how can this be further interpretated?

$\endgroup$

1 Answer 1

1
$\begingroup$

To be honest, I'd be approaching this from the view of the theory I was exploring, rather than just hunting through the data for something interesting. I'd certainly expect EYE_WIDTH_MAX and EYE_WIDTH_AVG to be very highly correlated. So I would have decided in advance which to consider.

Secondly, I apologise if I'm missing some context here, but why do you have MAX and AVG in here. Why not include all observations (for individuals) and have a random effect for each individual?

Thirdly, I think you need to think hard which variables you want to include - do you want Start and Stop timestamp for example, or do you want the difference between the two (length of procedure??????).

It looks to me if some variables are perfectly correlated. So one thing you could do is look at a pairwise scatterplot of your raw data before you do any modelling. There's a lot of really good ways of approaching visualisation and modelling. I'm guessing some of your variables are categorical, but again good visualisation and tabulation will help here.

$\endgroup$
1
  • $\begingroup$ Great! Thank you for your insight. I have done much analysing and visualization. But this was just considered to see amount of correlation within data, so all variables were included. But those variables with AVG and MAX, MIn or timestamps are not considered. So far what I read if VIF factor is more than 5, it is highly correlated.But here I got so much larger number which makes my interpretabilty difficult. But also I am finding it difficult how can I model these data in which few variables are high cardinal few are low cardinal categorical variables and some are numerical ,this is classifier $\endgroup$
    – Chinti
    Mar 4, 2020 at 14:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.