1
$\begingroup$

If I have variables $X_1$, $X_2$, ... $X_n$ whose values are only 1 or 0, in order to explain a variable $Y$, is there a way to determine which ones are significant, and also, how could it be measured?

$\endgroup$
  • $\begingroup$ This is a regression problem. The method to select depends on the nature of $Y$ and, to some degree, on how frequently the various combinations of the $X_i$ appear in the data and precisely how you view $Y$ as depending on the $X_i$. If you could disclose that information you would have much better chances of attracting answers that go beyond the generic advice "use regression" or "use ANOVA" (which is essentially the same thing). $\endgroup$ – whuber Jun 20 '17 at 19:36
0
$\begingroup$

The simplist approach is an ANOVA using a model with no interaction terms (leave in the interaction terms if you want to test them).

$\endgroup$
  • $\begingroup$ I tried fitting a logistic regression model for a simple case, which consisted of a variable that contained values of 0 and 1 and another variable (dependent) that is a copy of the other one. I expected the independent variable to be significant, but in the summary the p-value for the Wald test was extremely high (equal to 1), indicating that the independent variable was not significant. What could be the problem? $\endgroup$ – Alfie González Jun 23 '17 at 16:01

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.