Bioequivalence and Bioavailability Forum

Hi Bebac user,

❝ In studies that have two tests (T1 T2) and one reference (R1)

❝

❝ If I treated them in statistical analysis as on part ---> it called " William's design"

Depends. If the designs had six sequences (T1T2R|T1RT2|T2T1R|T2RT1|RT1T2|RT2T1), it was a Williams’ design. If it had three sequences (T1T2R|T2RT1|RT1T2), it was a Latin Square.

❝ But if I treated them in statistical analysis as two parts

❝ (T1 vs R and T2 vs R) ---> it called "Latin square design"

❝

❝ IS THIS RIGHT?

No. Design ≠ evaluation. You can evaluate a Williams’ design either by a pooled ANOVA (that’s the ‘All-at-Once’ approach) or as two incomplete block designs (that’s the ‘Two-at-a-Time’ approach).
See this article.