Discrepancies in p-values between t.test vs regression after random forest analysis

by MAPK   Last Updated August 14, 2019 17:19 PM

I was trying to analyze two groups of samples for multiple variables. I first used Boruta (random forest analysis) test to determine the importance of variables in my data. Boruta test Boruta analysis

Boruta test identified Actinomycetes% to be the only (significantly) important variable. Then I did t.test and regression analysis on the variable Actinomycetes% and found them to be not significant. Also, the p-values I got from t.test is ~0.14 and regression analsysis is ~0.9. So, I would like to ask the community these two questions:

  1. Why does Boruta test show Actinomycetes% to be significant and not by t.test and regression analsysis?
  2. Why are the p-values from t.test (0.19) substantially smaller than the p-value from regression analysis(0.9)?

t.test T-test

regression analysis Regression analysis

Related Questions

Boruta score goes to minus infinity

Updated May 23, 2018 07:19 AM

Boruta Rejecting Important Variables

Updated September 07, 2018 11:19 AM

What does z-score mean in Boruta

Updated May 24, 2018 13:19 PM