Slides feedback

[mvanrongen]

Rob:

• Slide 128/133/142 - I would change "model" here to either "fitted values", "fit", or "prediction". This is because the residuals are a part of the model - the model comprises both a predicted (average) value as well as uncertainties associated with the values about that prediction. I always try to drive it in that the "model" isn't just the line of best fit!
• Slide 207 - After identifying that the interaction term is not significant, it's important to create a new model without the interaction term before interpreting the main effects. It won't change the results much but it will change the results a bit, especially when there are multiple groups per predictor, due to the effect on the residual degrees of freedom.