I need to make a residual plot and I was wondering whether I make the plots in multiple linear regression on one independent variable at a time (like making a simple linear regression) or the all of the ten independent variables at the same time (like multiple linear regression)? They produce different results for me obviously.
$\begingroup$
$\endgroup$
3
-
1$\begingroup$ Hello - "residual plot" can refer to many different things. What is your goal? Also you may want to look into partial plots, a.k.a. partial regression plots. $\endgroup$– rolando2Apr 27, 2020 at 15:17
-
$\begingroup$ My goal is to check heteroscadisticity and linearity of the data. I have ten independent variables and I'm not sure whether to plot the residuals individually against dependent variable or all of them at the same time, like when doing a multiple linear regression $\endgroup$– Olivia HeinoApr 27, 2020 at 15:22
-
$\begingroup$ Each plot is valuable, and in addition you should inspect fitted values versus residuals. But no finite amount of plots will be guaranteed to "catch" heteroscedasticity or non-linearity if it exists. $\endgroup$– AdamOApr 27, 2020 at 15:45
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
To check for overall heteroscedasticity:
- On the Y-axis: your model's residuals
- On the X-axis: either your dependent variable or your predicted value for it. You might try a plot using each.
Note that John Fox in Regression Diagnostics finds that, typically, only when the variance of the residuals varies by a factor of three or more is it a serious problem for regression estimation.
To check for overall linearity:
- On the Y-axis: your dependent variable
- On the X-axis: your predicted value for the dependent variable
Then you might create a linear fitline and one using a lowess and/or a quadratic or even a cubic fit, to compare to the linear one.
To check for heteroscedasticity, linearity, and influential points with respect to each X-Y relationship:
- Create partial plots, a.k.a. partial regression plots. Each will show an individual X-Y relationship while controlling for the other predictors.
