SCIENCE ADVANCES | RESEARCH ARTICLE
Statistical analyses
Responses to transfer trials were identified and organized by the response types. Only trials in which all six stimuli were chosen were used in the following analyses; we excluded miss and incomplete trials. Miss trials occurred when the bracket stimuli appeared on the screen, but the crow did not response to any stimulus. Incomplete trials when the bird responded to at least one but not all stimuli on the screen.
All analyses were done in MATLAB version 2018,2020 and R. We used binomial tests to test whether different responses were different from chance. To test whether there was a difference in the frequency of responses, we used chi-square tests for independence and goodness of fit with Yates' continuity correction. All tests were two-tailed. Numbers, test statistics, and P values are reported. Alpha was set to α = 0.05.
We generated a response matrix for each transfer trial list to examine the relationship between bracket position and reliability of the responses. Because each column represents a position in the sequence, we isolated the highest proportion in each column and then calculated a correlation coefficient (MATLAB function: corrcoef). We got a total of 36 coefficients (18 different transfer trial lists, two crows), and we tested these values against the null with a one-sample t test.
SUPPLEMENTARY MATERIALS
Supplementary material for this article is available at https://www.science.org/doi/10.1126/sciadv.abq3356
View/request a protocol for this paper from Bio-protocol.
