Identification of statistical tests used and parameters relevant to post hoc power analysis
Data | Test | Power (%) | Effect size | Noncentrality | n | |
---|---|---|---|---|---|---|
a' | Unknown | Paired t test | 9 | 0.28 | 0.70 | 6 |
b' | Unknown | Paired t test | 5 | 0.01 | 0.02 | 8 |
a | Normal | Paired t test | 70 | 0.79 | 2.73 | 12 |
b | Unknown | Paired t test | 39 | 0.53 | 1.85 | 12 |
c | Normal | 1 sample t test | 6 | 0.07 | 0.25 | 12 |
d | Unknown | Wilcoxon rank | 91 | 2.11 | 3.57 | 4,12 |
e | Unknown | Wilcoxon signed rank | 5 | 0.11 | 0.21 | 4 |
f | Unknown | Wilcoxon rank | 9 | 0.39 | 0.65 | 4,12 |
G | Unknown | Wilcoxon rank | 81 | 1.81 | 3.07 | 4,12 |
H | Unknown | Wilcoxon rank | 6 | 0.27 | 0.35 | 3,4 |
I | Unknown | Wilcoxon rank | 9 | 0.58 | 0.74 | 3,4 |
J | Unknown | Wilcoxon rank | 5 | 0.006 | 0.007 | 3,4 |
K | Normal | 1 sample t test | 100 | 1.73 | 6.47 | 14 |
L | Unknown | Mann–Whitney–Wilcoxon | 8 | 0.36 | 0.59 | 4,10 |
M | Unknown | Paired t test | 100 | 1.52 | 5.56 | 14 |
N | Unknown | Paired t test | 15 | 0.27 | 1.0 | 14 |
P | Normal | Paired t test | 97 | 1.11 | 4.14 | 14 |
Q | Unknown | Paired t test | 6 | 0.07 | 0.26 | 14 |
R | Unknown | Paired t test | 10 | 0.19 | 0.71 | 14 |
S | Normal | Paired t test | 80 | 1.07 | 3.21 | 9 |
T | Unknown | Paired t test | 5 | 0.03 | 0.10 | 9 |
U | Unknown | Paired t test | 5 | 0.06 | 0.17 | 8 |
V | Unknown | Paired t test | 5 | 0.06 | 0.18 | 8 |
W | Unknown | Paired t test | 1 | 4.86 | 10.86 | 5 |
X | Unknown | Paired t test | 48 | 1.13 | 2.52 | 5 |
Y | Unknown | Paired t test | 100 | 11.48 | 28.11 | 6 |
Z | Unknown | Paired t test | 14 | 0.43 | 1.06 | 6 |
y' | Unknown | Paired t test | 100 | 5.82 | 14.27 | 6 |
z' | Unknown | Paired t test | 47 | 0.96 | 2.34 | 6 |
Data structure was tested for normality when parametric tests were applied using Kolmogorov–Smirnoff tests, but was of limited application here due to the modest sample sizes inherent in behavioral studies (<20 animals/group). In all cases, nonparametric tests yielded the same results. Effect size and measures of predicted noncentrality are also given to properly infer cases in which post hoc calculations of statistical power were weak. In most cases of weak power the effect sizes were medium to large (>0.35). In cases for which the effect size was very small (<0.10) the predicted noncentrality measures were also very low (<0.30), indicating that any potentially real differences between groups was minimal such that additional sampling within reasonable limits would be unlikely to result in statistical significance. It should be noted that only substantial changes in behavior were of interest in the current work. Post hoc power analysis was performed using G*Power (v3.1.9.2, www.gpower.hhu.de).