Supplementary Information: A theoretical model for the associative nature of conference participation.

Proportion of conference participants g with x conference attendances who are going to attend the conference next year (Figure A). Proportion of conference participants ρ with n missed conferences after x-th conference attendance who are going to skip the conference next year, but will take part at some future conference from the observation period (Figure B). Pages on the web from which we downloaded conference data (Table A). Summary of the conference data. Columns 2 and 3 indicate for each conference the year in which data we have collected begin (Y0) and end (Yf). The total number of different participants at the conference during that period of time is given in column 4 (Table B). The number of participants at the conference per year (Table C). Log likelihood ratio and the π-value compare the fit to the power-law with the fit to the power-law with an exponential cutoff for the probability distribution of number of conferences at which each author appears (Table D). Log likelihood ratio and the π-value compare the fit to the power-law with the fit to the power-law with an exponential cutoff for the probability distribution of the number of successive participations at the conference (Table E). Log likelihood ratio and the π-value compare the fit to the exponential with the fit to the power-law with an exponential cutoff for the probability distribution of the time lag between two consecutive conference participations (Table F). The optimal parameter values of the model for each conference (Table G). Stagnancy rate 1 − g(1, 0) at t = 1 for each conference and exponent α of power-law with an exponential cutoff distribution fit with the corresponding conference order (Table H). (PDF)