11.Significance Testing for Fixed Effects

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:
Significance Testing for Fixed Effects
- Base on ‘Wald’ F and t tests
- Degrees of freedom (DF) for these tests should:
--- Correspond to the variance of contrast
--- Reflect all sources of variability
- Balanced (or complete) data :
--- F is ratio of mean squares
--- Denominator DF then corresponds to DF of denominator mean square (often the residual)
- Unbalanced data :
--- DF less straightforward to calculate
--- Satterthwaite’s approximation available

Hypertension Trial Example – Significance tests, SAS Output
- Test of overall treatment equality :
- Pairwise comparisons of treatments :
- Denominator DF varies due to the differences in distribution of treatments across centres.