0
$\begingroup$

I have a dataset of 300 observations with 7 predictor variables with 1 continuous response variable. The response is strongly skewed to the right and there are no significant correlations among any of the predictors or the response. I have attempted several different models including XGBoost, random forests, etc. Even transforming of the response hasn't helped. Any advice on how one should approach a problem like this would be greatly appreciated.

Improve this question
$\endgroup$
3
  • $\begingroup$ It seems 7 predictors have no relation with response variable. Given there really is no relation, the statistical methods can not create the relation. $\endgroup$
    – user158565
    Commented Jul 13, 2019 at 23:38
  • $\begingroup$ What is the response measuring? Is it a count, an amount of money, the size of something, a concentration, an angle, ...? $\endgroup$
    – Glen_b
    Commented Jul 14, 2019 at 0:34
  • $\begingroup$ Sorry didn't clarify that. The response is proportions between 0 and 1 with most values falling close to 0 and the max being around .14. $\endgroup$
    – Jeff Stats
    Commented Jul 14, 2019 at 0:59

0

Reset to default