A residual is what the model failed to explain for one data point: observed minus fitted. Plotting residuals is the main check on any regression or ANOVA. Against fitted values they should form a shapeless band; a funnel means variance grows with the mean, and a curve means a missing term. Against run order they should show no trend, and on a normal probability plot they should fall near a straight line.
Residuals fan out above 300 nm, so the error isn't constant. Try a log transform before you trust those p-values.
Always plot residuals against run order: a slope there means drift during the experiment, not a factor effect.
Also heard as
- residuals
- residual error
Related terms
Regression
Fitting an equation that predicts a response from one or more input variables, usually by least squares.
Outlier
Observation that sits far from the pattern of the rest of the data, either a genuine extreme or an error.
Normal distribution
Symmetric bell-shaped distribution set by a mean and a standard deviation, the default model for measurement data.
ANOVA
Analysis of variance: splitting total variation into parts from each factor and from error, then testing which parts are real.