57ª Reunião Anual da Rbras Workshop in Design of Experiments (DOE) Analysis for Repeated Crossover Designs Carla A. Vivacqua André L. S. de Pinho Federal University of Rio Grande do Norte UFRN – Brazil Outline • Motivation • Good milking practices study • Experimental Design • Double-reversal crossover design • Analysis • Future Work 3 Motivation • Safety of dairy products: • Human health • Environment • Animal well-being 4 Somatic Cells Count • Somatic cells count - SCC (Andrade et al., 2004): • Quality of milk • Health of mammal glands 5 Good Milking Practices • Set of guidelines for milking process: • • • • Cleaning of all equipments Hygiene of operators Cleaning of the environment Cleaning solutions used before and after each milking 6 Objective • Evaluate the influence of good milking practices on somatic cell count 7 Previous Studies • Other variables that may affect somatic cells count : • Parity order • Lactation period 8 Experimental Design Experimental Units and Treatments • Experimental units: • Holstein crossbred cows • Treatments : • Milking using good practices guidelines • Milking without good practices guideline (control) • Just washed the mammal glands with water 10 Completely Randomized Design 11 Randomized Block Design Grouped by parity order and lactation period 12 How homogeneous are the experimental units? Crossover Design • AB/BA design 14 Crossover Design • Longitudinal study in which experimental units receive different treatments at different periods of time, in a predetermined sequence • Objective: Reduce the effect of differences among experimental units on the comparison between treatments • Simplest crossover design (AB/BA): involves two treatments and two periods of time • Two concepts are fundamental: the time interval between applications of different treatments (washout) and residual effect of treatments (carryover). 15 Washout Period 16 Case Study • Washout: - Ideally, the time interval should be large enough so that the residual effects of the previous treatment is eliminated; • Milking good practices study: - Washout period of 7 days 17 Carryover Effect • Residual effect of a treatment • Treatment effect may persist a long time after its application • Main concern during the analysis of a crossover design Kenward, 2003) 18 (Jones and Double-Reversal Design • Four periods of time (four weeks) • Two sequences of treatments: • ABAB • BABA • 16 Holstein crossbred cows • Grouped by: • Parity order • Lactation period • Four cows in each group 19 Analysis Linear Model • Terms: • • • • Intercept Period effect Y Direct treatment effect Carryover (additive and treatment by carryover interactions) • Effect associated with an animal on a sequence (parity order, lactation period effects and interactions) 21 Basic Command in R • Specify the correct error terms • Error 22 Future Work • During the execution of the experiment: • One cow got sick before the start of the experiment • Delay in the start of the experiment led to the classification of a cow on a different lactation period • One cow got sick during the experiment, some observations periods are missing • How do these affect the analysis? • Would another design be better? 23 References • ANDRADE et al. (2004) “Influência da contagem de células somáticas sobre a produção de leite em diferentes fases da lactação”, V Simpósio da Sociedade Brasileira de Melhoramento Animal • JONES, B., KENWARD, M.G. Design and Analysis of Cross-Over Trials, Chapman & Hall/CRC, New York, 2003. 24 Carla A. Vivacqua [email protected] André L. S. de Pinho [email protected] Thank you!