Скачать или смотреть 12.Model Assumptions & Model Checking Methods

Introduction to Mixed Models
With Dr Helen Brown, Senior Statistician at The Roslin Institute, March 2016
*Recommended Youtube playback settings for the best viewing experience: 1080p HD
************************************************
Content:
Model Assumptions
Residuals and random effects should be :
--- Normally distributed
--- Uncorrelated
Consequences of violating assumptions not fully known
Model checking methods not developed in depth

Simple Visual Checks
Detect important deviations from assumptions
Carry out for residuals and random effects
1. Residual plots
--- Plot residuals (or random effects) against their predicted values
--- Rough check of normality
--- Check whether variance is constant across range of values
--- Spot outliers
2. Normal probability plots
--- Plot of residuals (or random effects) against their predicted values given their ranks
--- Further check of normality
3. Compare residual (or random effect) variance between groups, eg treatments
--- Check homogeneity of variances

Residual plots and statistics (SAS output)

Checking Normality of Centre Effects