BIZNIQUE / LINEAR ALGEBRA / 37
Look at what the model leaves behind
Look at what the model leaves behind
Residual structure can reveal a problem that a single error score hides.
Embedded Files
THE IDEA
Read the example.
Read the example.
Each vertical segment connects an observation to its fitted value. Squaring and averaging residuals produces mean squared error. Patterns across x, time or groups can suggest missing structure, changing variance or unsuitable assumptions.
Residual = observed − predicted
Use it well
Use it well
Small residuals on training data do not establish generalisation. A residual pattern is a diagnostic clue, not a complete explanation of its cause.
TRY IT YOURSELF
A quick check.
A quick check.
An observation is 7 and the prediction is 5. What is its residual and squared residual?
Check your reasoning
Check your reasoning
The residual is 2 and its square is 4.
Page updated
Google Sites
Report abuse