Details of statistical treatments.
| Line | Hypothesis (H0) | Test | df | Test statistic value | Probability | Outcome | Power of outcome | Sample type | Subjects excluded (n) | Reason for exclusion |
|---|---|---|---|---|---|---|---|---|---|---|
| a | Mean of lose-shift probability across the population is not equal to 0.5. | t | 97 | 19.2 | 1.00E–34 | Reject H0 | 1 | Subjects | 0 | |
| b | Mean of win-stay probability across the population is not equal to 0.5. | t | 97 | 1.4 | 0.17 | Accept H0 | 0.74 | Subjects | 0 | |
| c | Relationship between win-stay and lose-shift across subjects is not linearly correlated. | Linear regression | 97 | 32.2 | 1.00E–06 | Reject H0 | 0.72 | Subjects | 0 | |
| d | Relationship between lose-shift probability and ITI computed from binned aggregate data from all subjects is explained by a constant model. | F vs. constant model | 14 | 398 | 1.00E–11 | Reject H0 | 1 | Binned probabilities | 0 | |
| e | Mean regression slope computed from the independent log-linear regression of lose-shift to ITI is not different from 0. | t | 54 | 40 | 1.00E–40 | Reject H0 | 1 | Subjects | 42 | Insufficient samples for regression (criterion is ≥25 samples in 4 consecutive bins, after removing trials that follow entry of the non-chosen feeder) |
| f | Relationship between win-stay probability and ITI for binned data across subjects is explained by a constant model. | F vs. constant model | 14 | 12.8 | 1.00E–03 | Reject H0 | 0.99 | Binned probabilities | 0 | |
| g | Mean regression factor for the quadratic term computed from the independent regression of lose-shift to log10(ITI) is not different from 0. | t | 63 | 6.6 | 1.00E–08 | Reject H0 | 0.96 | Subjects | 32 | Insufficient samples for regression (criterion is ≥25 samples in 4 consecutive bins, after removing trials that follow entry of the non-chosen feeder) |
| h | Relationship between the ITI after wins and the ITI after losses is explained by a constant model. | F vs. constant model | 97 | 225 | 1.00E–26 | Reject H0 | 1 | Subjects | 0 | |
| i | Relationship between subject-wise lose-shift probability and logarithm of the ITI after losses is explained by a constant model. | F vs. constant model | 97 | 20.6 | 2.00E–05 | Reject H0 | 0.99 | Subjects | 0 | |
| j | Relationship between subject-wise win-stay probability and logarithm of the ITI after wins is explained by a constant model. | F vs. constant model | 97 | 1.8 | 0.18 | Accept H0 | 0.6 | Subjects | 0 | |
| k | Response time is invariant to the trial position within sessions, independent of barrier length (i.e., main effect). | RM-ANOVA | 9,864 | 2.8 | 0.003 | Reject H0 | 0.96 | Binned trials and subjects | 0 | |
| l | Anticipatory licking is invariant to the trial position within sessions, independent of barrier length (i.e., main effect). | RM-ANOVA | 9,864 | 8.8 | 1.00E–06 | Reject H0 | 1 | Binned trials and subjects | 0 | |
| m | Relationship between the within-session change in anticipatory licking and total licks (per trial) is explained by a constant model. | F vs. constant model | 8 | 38.7 | 3.00E–04 | Reject H0 | 0.99 | Binned trials | 0 | |
| n | The prevalence of lose-shift responding is invariant to the trial position within sessions, independent of barrier length (i.e., main effect). | RM-ANOVA | 9,864 | 2.2 | 0.02 | Reject H0 | 0.89 | Binned trials and subjects | 0 | |
| o | Relationship between the within-session change in lose-shift prevalence and anticipatory licking is explained by a constant model. | F vs. constant model | 8 | 27.8 | 7.00E–04 | Reject H0 | 0.99 | Binned trials | 0 | |
| p | ITI after loss is invariant to the trial position within sessions, independent of barrier length (i.e., main effect). | RM-ANOVA | 9,864 | 29 | 1.00E–06 | Reject H0 | 1 | Binned trials and subjects | 0 | |
| q | Relationship between the within-session change in lose-shift prevalence and log ITI after loss is explained by a constant model. | F vs. constant model | 8 | 24.8 | 1.00E–03 | Reject H0 | 0.99 | Binned trials | 0 | |
| r | Mean running speed in the presence of shorter barriers is not different from the mean running speed in the presence of the longer barriers. | t | 18 | 0.05 | 0.96 | Accept H0 | 0.96 | Subjects | 0 | |
| s | Mean % change in A.U.C for lose-shift vs. log(ITI) due to increasing barrier length for each subject is not different from 0 | t | 16 | 0.09 | 0.93 | Accept H0 | 0.95 | Subjects (within) | 2 | Insufficient samples for regression (criterion is ≥25 samples in 4 bins) |
| t | Mean % change in A.U.C for win-stay vs. log(ITI) due to increasing barrier length for each subject is not different from 0 | t | 14 | 0.55 | 0.59 | Accept H0 | 0.87 | Subjects (within) | 5 | Insufficient samples for regression (criterion is ≥25 samples in 4 bins) |
| u | Mean change in lose-shift probability across subjects when the longer barrier is introduced is not different from 0. | t | 18 | 4.7 | 2.00E–04 | Reject H0 | 0.71 | Subjects (within) | 0 | |
| v | Mean difference between predicted and actual lose-shift decrease due to increased barrier length is not different from 0. | t | 18 | 0.14 | 0.89 | Accept H0 | 0.95 | Subjects (within) | 0 | |
| w | Mean change in rewarded trials due to barrier length is not different from 0. | t | 18 | 2.45 | 0.02 | Reject H0 | 0.92 | Subjects (within) | 0 | |
| x | The prevalence of lose-shift responding is invariant to the trial position within sessions, independent of barrier length (i.e., main effect). | RM-ANOVA | 6,109 | 1.6 | 0.16 | Accept H0 | 0.42 | Binned trials and subjects | 0 | |
| y | The ITI after loss is invariant to the trial position within sessions, independent of barrier length (i.e., main effect). | RM-ANOVA | 6,109 | 5.7 | 3.00E–05 | Reject H0 | 0.99 | Binned trials and subjects | 0 | |
| z | Anticipatory licking is invariant to the trial position within sessions, independent of barrier length (i.e., main effect). | RM-ANOVA | 6,109 | 6.8 | 4.00E–06 | Reject H0 | 1 | Binned trials and subjects | 0 | |
| aa | The prevalence of lose-shift responding is invariant to barrier length, independent of changes due to trial position in the session (i.e., main effect). | RM-ANOVA | 1,18 | 8.3 | 0.01 | Reject H0 | 0.78 | Binned trials and subjects | 0 | |
| ab | The ITI after loss is invariant to barrier length, independent of changes due to trial position in the session (i.e., main effect). | RM-ANOVA | 1,18 | 28 | 5.00E–05 | Reject H0 | 1 | Binned trials and subjects | 0 | |
| ac | Anticipatory licking is invariant to barrier length, independent of changes due to trial position in the session (i.e., main effect). | RM-ANOVA | 1,18 | 0.5 | 0.52 | Accept H0 | 0.9 | Binned trials and subjects | 0 | |
| ad | Relationship between lose-shift responding and anticipatory licking is explained by a constant model. | F vs. constant model | 5 | 10.1 | 0.02 | Reject H0 | 0.58 | Binned trials | 0 | |
| ae | Mean difference in win-stay probability across subjects computed after a previous win vs. two previous wins at the same feeder is not greater than 0. | t | 48 | 10.2 | 1.00E–13 | Reject H0 | 1 | Subjects (within) | 2 | Insufficient occurrence of win-stay-wins sequences (criterion is ≥25) |
| af | Mean difference in lose-shift probability across subjects computed after a previous loss vs. two previous losses at the same feeder is not greater than 0. | t | 32 | 2.2 | 0.99 | Accept H0 | 1 | Subjects (within) | 18 | Insufficient occurrence of lose-stay-lose sequences (criterion is ≥25) |
| ag | Mean prediction accuracy of the Q-learning model and win-stay-lose-shift is not different from 0. | t | 34 | 5.2 | 1.00E–05 | Reject H0 | 0.96 | Subjects | 0 | |
| ah | The median probability of lose-shift on the second training session is not different from chance (0.5). | Wilcox | 17 | 0.03 | Reject H0 | 0.77 | Subjects | 0 | ||
| ai | Mean probability of lose-shift did not change across training or testing days. | RM-ANOVA | 15,150 | 0.54 | 0.91 | Accept H0 | 1 | Subjects, sessions | 0 | |
| aj | Mean probability of win-stay did not change across training or testing days. | Wilcox | 17 | 0.01 | Reject H0 | 0.83 | Subjects | 0 | ||
| ak | Mean probability of win-stay did not change across training or testing days. | RM-ANOVA | 15,150 | 2.3 | 5.00E–03 | Reject H0 | 1 | Subjects, sessions | 0 |