FAQ: Can we use iNZight to work out residuals and draw residual plots? (3.9)

I guess you are talking about residual plots for linear regression in which case you will have to form the residuals “by hand” using the Manipulate Variables menu

Suppose that you were predicting height from age:

  1. Drag variable names to produce the height vs age  scatter plot
  2. Use Add to Plot to put a linear trend on it
  3. Click Get Summary (or Get Inference) to obtain the intercept and slope from the fitted line. Suppose it turns out:  int=96.71 slope = 4.83
  4. Use “Create New Variables” which is under “Manipulate Variables”.
  5. For understandability I’d do it in 2 steps …
    a.  Create PredHeight = 96.71 + 4.83*age
    b.  Create Resids = height – PredHeight

Then you can plot Resids against anything you like  (in iNZight all variable names are case sensitive)

I didn’t get this “canned” in the basic part of iNZight because in a bivariate setting (a single predictor variable) you seldom see anything in a residual plot you can’t see in the original scatter plot so I would not stress residual plots in a first bite at regression (except maybe for your brighter students).

If you turn them loose on the Model Fitting module under “Advanced” then residual plots are generated automatically but that part was really designed for university courses.

  • (Last updated: 12/02/13. Added: 12/02/13)