Extended Data Fig. 4: Regression coefficients for research areas as a function of research methods.
From: Differences in psychologists’ cognitive traits are associated with scientific divides
Like Fig. 3, each cell here gives the regression coefficient for the simple model where individual research areas are regressed on individual research methods (though as the y-variable here is binary, unlike the controversial themes in Fig. 3, these are binomial regression coefficients expressed in logits). Cells marked with ‘x’ are non-significant (Bonferroni correction for multiple comparisons—the number of cells in the panel—yielding threshold p < .0001515). The margins of the plot show hierarchical clusters derived using Ward’s method. If the vectors of individual responses for research areas represent one high-dimensional space and the vectors of individual responses for research methods represent another high-dimensional space, then the correlation between all pairs of research-area vector cosine similarities and all pairs of research-methods cosine similarities is ρ = 0.22 (Figs. 5, 7), representing a conservative estimate for the overall association between these two types of response. For numeric results, see Appendix 2, Supplementary Table 6.
