Fig 5. Estimated probability density functions of p-values conditioned to p ≥ 0.01 separating for different ranges of text length L. p-values correspond to the fitting of word frequencies to (a) distribution f₁ and (b) distribution f₂.

We divide the distribution of text length into 15 intervals of 2 000 texts each. For distribution f₁ only the first seven groups (up to length 34 400) are displayed (beyond this value we do not have enough statistics to see the distribution of p-values greater than 0.01, as displayed in Fig 6; for distribution 2 this happens only in the last two groups). The intervals L_i range from L₁ = [115, 5291] to L₆ = [25739, 34378] and to L₁₃ = [89476, 103767].