0

After running my regression, none of my variables are significant so I was wondering if conparing t and p values could be a source of discussion instead... If so, can anything be inferred when a t value is larger than a p value? Or the other way around? Or what does it mean when the p and t values are almost equal?

LLGG
  • 13
  • 1
    Welcome to Cross Validated! 1) This makes little sense to me. What do you hope to accomplish? 1a) Why are you trying to force some kind of model significance? 2) What kind of overall p-value do you get for your model? Particularly if your variables are related, it is common to have no significant variables yet a highly significant model. Being a bit loose, I interpret this to mean that the regression cannot figure out which variable causes what, but it knows that the variables together do a lot. – Dave Mar 22 '23 at 00:01
  • 1
    Comparing a t-statistic to a p-value is no more meaningful than comparing a weight to a height. https://stats.stackexchange.com/questions/31 might be a good place to begin learning more about what these quantities are and how to interpret them. – whuber Mar 22 '23 at 00:07
  • 1
    p and t are related through the simple formula $p = 2\cdot (1-(\mbox{pt}(t, \mbox{df}))$, so I guess you do not want to compare t with p, but mean something different. Maybe comparing the t values of different variables (which are often used as an index for "variable importance")? Are the t-values for all variables almost equal? – cdalitz Mar 22 '23 at 10:36

0 Answers0