Men deny more than they believe about climate change on Twitter (X)

large dataset (millions of tweets). Conceptually, the model selection is also based on the assumption that no one is considered superior in holding their stance towards climate change and hence the stances cannot be ordered. Before running the model, I omitted the year 2006 as it contained only four valid observa- tions after dropping the bank cells and also dropped the ‘undefined’ category of gender. After the cleaning, the values in data were left with the year 2006 to 2017. A descriptive summary of variables used in the analysis is given in Table 1 ahead. I have used RStudio software and nnet package [39] to conduct multinomial regression. The table indicates low number of cases in two categories (ideological positioning on global warming and impact of resource over consumption) of variable ‘Topics’. In all the other vari- ables, at least 10% cases are present within a variable. The % distribution of cases across the variable categories suggest sufficient number of cases to run the multinomial model with no blank cells as those were dropped from the data in cleaning process. The first category of each variable was held as referent to interpret the results. The model converged in 20 iterations and the significant chi-square test (33936, P<0.00) shows the statistical significant of the model [40]. The results are discussed in the following section. Results Table 2 shows the results of the model. While the sign of the coefficients tells the nature of comparative differences or similarity of opinions between the deniers and believer groups; the

log-odds interpretation and relative-risk ratio calculation further helps understanding the size of differences in opinions of the two groups across these variables. Table 1. Displays the descriptive of key variables summary of the categorical variables used in the model. Variables Categories Frequency % of N Stance Neutral 343751 32.78% Believer 595818 56.82% Denier 109006 10.40% Gender Female 310124 29.58% Male 738451 70.42% Topics Ideological Positions on Global Warming 74106 7.07% Impact of Resource Overconsumption 64975 6.20% Importance of Human Intervention 320802 30.59% Seriousness of Gas Emissions 139207 13.28% Weather Extremes 449485 42.87% Aggression Aggressive 322527 30.76% not aggressive 726048 69.24% Total number of cases N = 1048575 https://doi.org/10.1371/journal.pone.0303007.t001 Table 2. Results of multinomial regression. (Intercept) temp Impact of Resource Overconsumption Importance of Human Intervention Seriousness of Gas Emissions Weather Extremes not aggressive male believer 0.44(0.01) 0.02 (0.00) 0.16(0.01) 0.95(0.01) 0.55(0.01) -0.3(0.01) -0.05(0.00) -0.14 (0.00) denier -0.64(0.01) -0.04 (0.00) -0.85(0.02) -0.92(0.01) 0.05(0.01) -0.32(0.01) -0.51(0.01) 0.27 (0.01) neutral Reference category (N = 1048576, Chi square = 33936, p = 0.000). https://doi.org/10.1371/journal.pone.0303007.t002 PLOS ONE Men deny climate change on Twitter PLOS ONE | https://doi.org/10.1371/journal.pone.0303007 February 13, 2025 5 / 9