In my experiment I measure accuracy (0-1) in two groups, on two different types of stimuli. I use the generalized linear mixed model because I think this is the only model that can deal with binary DVs and with nested data (trials within the same participant).
GENLINMIXED /DATA_STRUCTURE SUBJECTS=id REPEATED_MEASURES=trial COVARIANCE_TYPE=AR1 /FIELDS TARGET=accuracy TRIALS=NONE OFFSET=NONE /TARGET_OPTIONS DISTRIBUTION=BINOMIAL LINK=LOGIT /FIXED EFFECTS=group stimulusfactorA stimulusfactorB group*stimulusfactorA group*stimulusfactorBstimulusfactorA*stimulusfactorBgroup*stimulusfactorA*stimulusfactorB USE_INTERCEPT=TRUE /RANDOM USE_INTERCEPT=TRUE COVARIANCE_TYPE=VARIANCE_COMPONENTS /BUILD_OPTIONS TARGET_CATEGORY_ORDER=ASCENDING INPUTS_CATEGORY_ORDER=ASCENDING MAX_ITERATIONS=100 CONFIDENCE_LEVEL=95 DF_METHOD=RESIDUAL COVB=MODEL /EMMEANS_OPTIONS SCALE=ORIGINAL PADJUST=LSD /SAVE PREDICTED_VALUES(PredictedValuere) PREDICTED_PROBABILITY(PredictedProbabilityre) MAX_CATEGORIES(25) CONFIDENCE(Confidence1re) CONFIDENCE_METHOD(PREDICTEDPROBABILITY) CONFIDENCE_INTERVALS(CIre) PEARSON_RESIDUALS(PearsonResidual1re).
In the fixed effect table with the F-tests, I see a very large effect of Stimulusfactor B (which has 4 categories). However, when I look at the simple contrasts, I see no significant effects at all. The graphs don't show the P-value, but just a red line (instead of a gold one), indicating that it is not significant. I feel something is wrong with this model but I am not sure what it is. Can someone help out?
PS When I analyze the averages per condition (percentages correct) in a simple repeated measures ANOVA, the pairwise comparisons do show significant differences.