Summary of statistical tests
| Data structure | Type of test | Result | Effect size | Power | |
|---|---|---|---|---|---|
| a | Categorical (binomial) | Fisher’s exact test | p = 2.8518e-06 | Odds ratio = 40.5 | 95% CI [6.57,249.65] |
| b | Leech ID: categorical (nominal) Mean normalized firing rate during response period: normally distributed; W(21) = 0.9469; p = 0.2353 | One-way ANOVA | F(9,12) = 0.7406, p = 0.6686 | η2 = 0.3571 | 0.1091 |
| c | Leech ID: categorical (nominal) Normalized mean absolute difference from baseline firing rate: normally distributed; W(21) = 0.9279; p = 0.1109 | One-way ANOVA | F(9,12) = 2.2830, p = 0.0918 | η2 = 0.6313 | 0.2794 |
| d | Mean baseline firing rate: non-normal; W(21) = 0.8446; p = 0.0040 Mean normalized firing rate during response period: normally distributed; W(21) = 0.9469; p = 0.2353 | Pearson’s correlation | p = 0.2067 | r = 0.2801 | 95% CI [−0.1604,0.6276] |
| e | Mean baseline firing rate: non-normal; W(21) = 0.8446; p = 0.0040 Normalized mean absolute difference from baseline firing rate: normally distributed; W(21) = 0.9279; p = 0.1109 | Pearson’s correlation | p = 0.0408 | r = −0.4261 | 95% CI [−0.7186,−0.0055] |
| f | Mean baseline firing rate in normal saline: normally distributed; W(12) = 0.9104; p = 0.1856 Mean baseline firing rate in Ca2+-free saline: non-normal; W(8) = 0.6781; p = 0.0024 | Wilcoxon rank-sum test | p = 0.2164 | r = 0.2635 | 0.2100 |
| g | Categorical (binomial) | Fisher’s exact test | p = 0.436 | Odds ratio = 2.2500 | 95% CI [0.3874,13.0665] |
| h | Coefficient of variability of baseline firing rate in high-heat US trials: normally distributed; W(21) = 0.9315; p = 0.1317 Coefficient of variability of baseline firing rate in low-heat US trials: normally distributed; W(19) = 0.9456; p = 0.3046 | Welch’s t test | t(40) = 1.2403, p = 0.2221 | d = 0.1343 | 95% CI [−0.0530,0.2213] |
Letters (leftmost column) correspond to statistical tests as reported in Results. The data structure, test type, result, effect size, and statistical power of these tests are described. Where applicable, results of Shapiro–Wilk tests for normality of data are reported under data structure. Effect sizes for Fisher tests are reported as odds ratios. One-way ANOVA effect sizes are reported as η2, calculated as the between-groups sum of squares divided by the total sum of squares. Effect sizes for Pearson’s correlation are the correlation coefficients. The effect size for the Wilcoxon rank-sum test is calculated as the z statistic divided by the square root of the population size, and the effect size of the Welch’s t test was calculated as Cohen’s d with a correction for small sample sizes as described (Durlak, 2009). When applicable, power was reported as the 95% confidence interval (CI) or statistical power calculated post hoc with G*Power.