I am using SPSS to run linear regression with several predictors. In some cases, when I threw in some variables, SPSS will show the regression model with all the variables. But at the bottom, it also shows a table named "Excluded variables."
I am not sure what it means. I suspect it may be a detection of multicollinearity involving these variables. BUt what I don't understand is the VIFs in the table show that there is no multicollinearity issue. Furthermore, these variables are still included in the previous model. So what does the "excluded variables" mean??
$\begingroup$
$\endgroup$
3
-
1$\begingroup$ Are you running an automated search like stepwise. $\endgroup$– user54285Commented May 24, 2019 at 15:54
-
1$\begingroup$ Yes. What does it mean? $\endgroup$– GeminiCommented May 28, 2019 at 20:20
-
$\begingroup$ Stepwise excludes variables, its the point of using it as a form of variable selection. I do not know SPSS, but I assume if you are using stepwise it is telling you what variables were excluded. Generally it is a very bad idea to use this method. $\endgroup$– user54285Commented May 30, 2019 at 0:16
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
"Excluded variables" in this context are those predictor variables that were either not added to and/or not retained in the final model. That doesn't mean that they are not important, and certainly not that they are not part of a causal system driving the behavior of the outcome variable. It just means what it says--the algorithm did its thing, and in the end those candidate predictor variables were not included in the regression model.
$\begingroup$
$\endgroup$
4
The excluded variables don't bring "additional significant information" to the model. In binomial logistic regression, depending on the method (forward or backward stepwise), you insert variables or remove on stages, where each stage is assesment whether the variable brings in additional information to the model.
-
1$\begingroup$ Does your "binomial logistic regression" mean "stepwise regression"? If not, could you explain what this phrase means and how it is relevant to the question? $\endgroup$– whuber ♦Commented Jun 13, 2019 at 14:23
-
$\begingroup$ Sorry, just read again the linear regression you mentioned. But the method is the same. More on the topic: en.wikipedia.org/wiki/Stepwise_regression $\endgroup$– JulianCommented Jun 13, 2019 at 20:27
-
1$\begingroup$ It's a familiar topic here, Julian: see stepwise regression. My comment was concerned about the strong impression you leave that "binomial logistic regression" and "stepwise regression" are the same or equivalent procedures--which they are not. $\endgroup$– whuber ♦Commented Jun 13, 2019 at 21:23
-
$\begingroup$ whuber, the question has been answered then. And the procedures are complementary, not equivalent. $\endgroup$– JulianCommented Jun 14, 2019 at 10:08