HUERISTIC ASSIMILATION SIMULATION: STATISTICAL ANALYSIS OF HUMAN'S SMELLING ONE ANOTHER'S PITS AND CRACKS....
The results, taking into consideration Human's relatively weak sense of smell, compared to most other creatures, of the "Smell of Age: Perception and Discrimination of Body Odors of Different Ages" tests were first converted to percentage correct values to allow the use of inference statistics on the underlying binary scale. To assess whether discrimination performance within each odor category differed from chance (50%), we used individual one-sample Student’s t-tests of percentage correct discrimination values. Differences in perceptual ratings were analyzed using mixed, repeated-measurements ANOVAs, separately for intensity and pleasantness, using a 2 (participant sex) ×3 (donor age group) ×2 (donor sex) design. Participant sex was entered as a between-subject factor, and donor age group and donor sex were used as within-subject factors. Subsequent Bonferroni posthoc tests were performed to statistically assess differences beyond main effects to control for multiple statistical comparisons. Discrimination performance was converted to a percentage correct discriminations value to correct for dichotomization. To assess statistical differences between stimuli, we used repeated-measurements ANCOVA, organized structurally as described above for perceptual ratings, but with two main differences: First, because our initial analyses of the perceptual ratings demonstrated that the largest perceptual difference between body odor groups was perceived intensity and not perceived pleasantness, intensity ratings for all age and sex groups were entered into the model as covariates (a total of six) to remove variance that could be explained by the intensity ratings. Please note that we refrained from adding the 6 pleasantness ratings as covariance factors because the unduly conservative nature of such analyses (a total reduction of our statistical power with 30% [−12 df]) would create a large risk of producing false negative results and conclusions. Second, because participants’ performance scores demonstrated near-significance on Mauchly’s Test of Sphericity (p = .058), all statistical values were submitted to Greenhouse-Geiser correction to avoid false positive results due to a skewed distribution. Age Labeling and Implicit Age Categorization task data (the latter described in detail below) were analyzed using chi-square (χ2) contingency tests against a random sampling distribution created using Monte Carlo simulations (n = 1000) within the statistical program R.
Comments
Post a Comment