Latin square coding

Fig. 3. The influence of administration of 5 mg/kg PCMG i.p. on median times of trial I ( S.E.M., calculated on the basis of Latin-square analysis). Coding of procedures as in Fig. 2 A, Control B, PCMG 20 min before trial I C, PCMG immediately following trial I D, PCMG 20 min before trial II. Significant differences are indicated by braces and asterisks , P < 0.05 , P < 0.01 , P < 0.001. For the definitions of latency, choice time and running time see text.

Four formulations of a drug are to be studied with respect to bioequivalence by treating four subjects at four periods. A 4 X 4 Latin square is constructed as given in Table 4.8. The formulations are coded by A, B, C, and D. 

Samples are presented in monadic seqnence, coded with three-digit random numbers, following a balanced rotation order (Williams Latin sqnare design) to avoid presentation order and carry-over bias. Hence, best practice requires the use of experimental designs to minimize sample presentation order bias and within-participant randomization of CATA terms. Typically, both designs will be based on Wiliams Latin square designs bnt be different in order to reflect the actual number of samples and terms used. Designs shoirld be developed to take into accoimt the number of consumers in the study. 

---