I have variables A and B and a target variable Y. Independently, A and not correlated to Y (the P value is too high) and B is correlated to Y with statistical significance. When I put A and B in a multivariate regression model to predict Y, A became significant and B stayed significant. Can someone explain the intuition behind why this happened? Is this why a lot of people say not to pay attention to the p value of a variable when it's in a multivariable regression (heard it from a professor)?
Asked
Active
Viewed 24 times
0
-
1In essence, look also at the relationship between your predictors. A scatter plot matrix should help. – Nick Cox Jan 25 '17 at 12:11
-
2Picky point: a multivariate regression is, strict sense, one with two or more response variables. Multivariate and multivariable aren't synonyms. – Nick Cox Jan 25 '17 at 12:13