Introduction to Statistical Modelling
Training session with Dr Helen Brown, Senior Statistician, at The Roslin Institute,  December 2015.
************************************************
These training sessions were given to staff and research students at the Roslin Institute. The material is also used for the Animal Biosciences MSc course taught at the Institute.
************************************************
*Recommended Youtube playback settings for the best viewing experience:  1080p HD
************************************************
Content:
Assumptions for GLMs
-Errors (or residuals) assumed :-
---normally distributed
---to have similar variance over ranges of variables and predicted values
-Check using :
---Residual plot: Plot of residuals (errors) against predicted values
---Normal plot: Plot of residuals (errors) against expected values given their ranks

Residual plot and normal plot for mouse injection model
-Residual plot: Plot of Residuals (Errors) against Predicted values
-Predicted = Intercept + Strain + Injection + Strain*Injection
---Basic check of normality
---Checks for outliers
---Checks if variance differs over predicted values
-Normal plot: Plot of residuals against expected values given their ranks (labelled ‘quantiles’)
---Further check of normality

Assumptions not met?
-Possible actions :
---Transformation of outcome variable (eg take logs)
---Deletion of outliers if it can be justified (eg error in recording?)
----- If no justification, could compare analyses with and without outliers
---Use non-parametric techniques (eg Wilcoxon rank sum test, Kruskal-Wallis, Spearman rank correlation/regression)
----- More restricted