Table 1.

Classical ANOVA of log intensities for duplicated A-loop design component of Figure 2 for any particular gene using (1).

Source	SS*	d f †	Mean square	Expected mean square
Inoculate	SSA	v A	MSA = SSA/v A	σ E 2 + 1.5σ R(A·B) 2 + ${{\phi }_A}^{‡}$
Time	SSB	v B	MSB = SSB/v B	σ E 2 + 2σ R(A·B) 2 + 2σ S(B) 2 + ${\phi }_B$
Inoculate*time	SSAB	v AB	MSAB = SSAB/v AB	σ E 2 + 1.5σ R(A·B) 2 + ${\phi }_{AB}$
Dye	SSD	v D	MSD = SSD/v D	σ E 2 + ${\phi }_D$
Rep(inoculate*time)	SSR(AB)	v R(AB)	MSR(AB) = SSR(AB)/v R(AB)	σ E 2 + 1.5σ R(A·B) 2
Array(time)	SSS(B)	v S(B)	MSS(B) = SSS(B)/v S(B)	σ E 2 + 1.5σ S(B) 2
Error	SSE	v E	MSE = SSE/v E	σ E 2

*Sums of squares.

^†Degrees of freedom.

^‡ φ _X is the noncentrality parameter for factor X. For example, when ${\phi }_A=0$ , there are no overall mean inoculate differences such that inoculate and Rep(inoculate*time) have the same expected mean square and F _A = MS_A/MS_R(AB) is a random draw from an F distribution with v _A numerator and v _R(AB) denominator degrees of freedom.