The Standardized Residual Plots tab contains the following elements:
|
•
|
These show the distributions of the standardized residuals. The standardization here refers to dividing the residuals by the square root of the estimated residual variance. Shown are Q-Q Plots, Box Plots, Histograms, quantiles, and summary statistics. The plots are particularly useful in visually verifying assumptions behind the ANOVA fit, more specifically, approximate normality. The ideal Q-Q plot forms a straight 45-degree line and the histogram is bell-shaped. Severe deviations from this indicate that anova assumptions are not met and results are suspect. In this case you might consider a different model or transformation for the data and/or automatic filtering of outliers.
|
•
|
A series of Bivariate Fit plots.
|
These plot the standardized residuals on the y-axis versus corresponding predicted value on the x-axis. These plots are useful for checking the linearity assumption of the anova model. An ideal plot shows random scatter around the 0 horizontal line and the estimated smoothing spline is approximately horizontal. Severe deviations from this indicate that anova assumptions are not met and results are suspect. In this case you might consider a different model or transformation for the data and/or automatic filtering of outliers.