RL37 Is the "Grizzle Option" the Best Approach When There's a Sequence Effect in a 2X2 Cross-Over
*Veronica N Taylor, U.S. Food and Drug Administration 


In the presence of a significant sequence effect from a 2X2 cross-over experiment, Grizzle recommended simply analyzing the first period data and abiding by the conclusions drawn from those results. The following alternatives will be discussed: (1) ignoring the sequence effect if all second period pre-treatment values are zero; (2) ignoring the sequence effect and adjusting for second period, non-zero, pre-treatment values; (3) using the treatment means estimated for the 1st period and the variance estimate from the cross-over analysis; and (4) not testing for sequence effect.