0
$\begingroup$

While building a model, when I add a function or an interaction effect there is an increase in my adjusted R2 value, However the term that I have added is not significant. In this case, can i still have the term or should i omit that term.

$\endgroup$

migrated from stackoverflow.com Oct 26 '15 at 0:23

This question came from our site for professional and enthusiast programmers.

  • 2
    $\begingroup$ More of a stats question. Even if it does get migrated as I am suggesting it would still be improves with more specific details of what is being done and the results. $\endgroup$ – DWin Oct 14 '15 at 23:32
  • 1
    $\begingroup$ agree with @BondedDust: you're going to have to give more detail. Can you actually give the adjusted R^2 values for the smaller and larger models, estimated parameter values/std err/t-statistic/df/p-value? Are you trying to test hypotheses or test models? $\endgroup$ – Ben Bolker Oct 14 '15 at 23:47
  • 1
    $\begingroup$ What are you trying to achieve? You certainly can have the term; nothing should stop you if you were inclined to include it to begin with. On the other hand omitting the term from the model on the basis of its lack of significance may not be as good an idea as you might have been led to believe. We have many answers here on CrossValidated that discuss the problems (including bias in parameter estimation of remaining terms, as well as bias in estimated parameter variance, bias in p-values, and so on) with using hypothesis tests (or indeed numerous other criteria) as a basis for selecting models $\endgroup$ – Glen_b Oct 26 '15 at 1:26

Your Answer

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