230 likes | 351 Views
Correlated-Samples ANOVA. The Multivariate Approach One-Way. Cross-Species-Fostering. House mice onto house mice, prairie deer mice, or domestic Norway rats. After weaning, tested in apparatus with access to tunnels scented like clean pine shavings, house mouse, deer mouse, or rat.
E N D
Correlated-Samples ANOVA The Multivariate Approach One-Way
Cross-Species-Fostering • House mice onto house mice, prairie deer mice, or domestic Norway rats. • After weaning, tested in apparatus with access to tunnels scented like clean pine shavings, house mouse, deer mouse, or rat. • House mice and deer mice were descendants of recently wild-trapped mice. • Reversed light cycle, red lighting
dataMus; infile'C:\ ... \tunnel4b.dat'; INPUT NURS V_clean V_Mus V_Pero V_Rat VT_clean VT_Mus VT_Pero VT_Rat T_cleanT_MusT_PeroT_RatTT_cleanTT_MusTT_PeroTT_Rat L_cleanL_MusL_PeroL_RatLT_cleanLT_MusLT_PeroLT_Rat; Format NURS rodent. ; The TT_ variables have been transformed to normal.
The ANOVA ProcANOVA; ModelTT_cleanTT_musTT_peroTT_rat = / nouni; Repeatedscent 4Contrast(1) / summaryprinte; run; • “nouni” suppresses irrelevant output • “summary” and “printe” gives us ANOVA tables for contrasts and “printe” tests sphericity
Contrasts • Contrast(1) – compare the first condition with all other conditions. • Profile – compare each condition with the next condition • Polynomial – trend analysis • Helmert – contrast each condition with the mean of the following conditions • Mean(n) -- contrast each level (except the nth) with the mean of all other levels.
Mauchly • Sphericity Assumption Violated
Univariate Approach • Both the G-G and the H-F are near or above .75, it is probably best to use the H-F • df = 3(.8422), 105(.8422) = 2.53, 88.43
Contrasts: Clean Scent vs. • Musmusculus: p = .008 • Peromyscusmaniculatus: p = .29 • Rattusnorvegicus: p = .14
Untransformed Means procmeans; varT_clean -- T_Rat;
Randomized Blocks Data data multi; input block1-block3; subj = _N_; B1vsB3 = block1-block3; B1vsB2 = block1-block2; B2vsB3=block2-block3; cards; 10 9 7 8 6 3 7 6 4 5 6 3 And two more cases
Randomized Blocks ANOVA • ProcANOVA; • Modelblock1-block3 = / nouni; • Repeatedblock 3 / nom;
Pairwise Comparisons procmeanstprt; var B1vsB3 B1vsB2 B2vsB3; run;
Want Pooled Error? • The comparisons on previous slide use individual error terms. • Get more power with pooled error. • First, unpack data from multivariate setup to univariate setup. • Then use ANOVA with desired procedure (LSD, Tukey, REGWQ, etc.)
Unpack the Data datauniv; set multi; array b[3] block1-block3; do block = 1 to 3; errors = b[block]; output; end; drop block1-block3;
The Unpacked Data And so on
LSD with Pooled Error ProcANOVA; Classsubj block; Modelerrors = subj block; Meansblock / lsdlines; run;
SPSS • Want to use SPSS instead of SAS? • See my document The Multivariate Approach to the One-Way Repeated Measures ANOVA