o ch 13 = r=1, factorial 2 level per treatment should block called 2**p - ex: BEADS A(size): S,L (0,1) (-1,1) ( ,a) B(color): B,O C(music): N,Y also know need to block AM,PM - what do treatment combos look like: 000 OR -1 -1 -1 OR (1) 001 OR -1 -1 1 OR (c) ... - if want full factorial with block Source df Block blk - 1 A a-1 B b-1 C c-1 A*B (a-1)*(b-1) A*C (a-1)*(c-1) B*C (b-1)*(c-1) A*B*C (a-1)*(b-1)*(c-1) Error Total n - 1 * orthogonal contrasts (like trend A2) won’t even work on this small sample - solution to our troubles? choose to confound higher order interaction with block *DEFN 2 effects are confounded when their estimators are identical * EX: H0: AM-PM no help H0: no A*B*C - this confounding seems good say, plots look like: - how to do this using orthogonal contrasts note:2 level factor main effects contrast easy. rest by multiplication 0 0 0 -1 -1 -1 1 1 1 -1 70.000000 0 0 1 -1 -1 1 1 -1 -1 1 65.000000 0 1 0 -1 1 -1 -1 1 -1 1 64.000000 0 1 1 -1 1 1 -1 -1 1 -1 62.000000 1 0 0 1 -1 -1 -1 -1 1 1 43.000000 1 0 1 1 -1 1 -1 1 -1 -1 41.000000 1 1 0 1 1 -1 1 -1 -1 -1 39.000000 1 1 1 1 1 1 1 1 1 1 35.000000 where to confound? block 1s and -1s together 000 001 010 011 100 101 110 111 BLOCK AM x x x x PM x x x x truly an incomplete block design like ch.11, for now analyze: - recall normalized orthogonal contrasts (independent) with df=1 sum of c(i)*Y(ijk) has variance (+-1)**2 sigma2 + ........(+-1)**2 sigma2 = v*sigma2 for all main and interaction (no divisor needed) Do usual sumc(i)Y(ijk) rnk blom cacee ------------- --- ---- ------ and plot SSC A 1326.125 1 1326.125 424.360 .031 B 45.125 1 45.125 14.440 .164 C 21.125 1 21.125 6.760 .234 AB .125 1 .125 .040 .874 AC .125 1 .125 .040 .874 BC .125 1 .125 .040 .874 ABC 3.125 1 3.125 Total 1395.875 7 199.411