Table 2.

Estimated LND parameters (μ, σ) for time duration distributions of phonemes, words and BG. Distribution candidates are fitted using MLE and model selection is based on maximizing the mean log-likelihood $⟨\mathcal{L}(\hat{\theta })⟩$. On the right side of the table, we depict Kolmogorov–Smirnov goodness-of-fit distance D_ks for different alternative distributions (data are more plausible to follow the distribution with lower values of D_ks). All tested distributions are defined with two parameters and inside each category level are evaluated in same conditions.

	LND		model selection $⟨\mathcal{L}(\hat{\theta })⟩$					goodness of fit Dks
	μ	σ	LND	beta	gamma	Weibull	normal	LND	beta	gamma	Weibull	normal
phoneme	− 2.68	0.59	1.8	1.76	1.76	1.69	1.48	0.014	0.052	0.05	0.081	0.128
word	− 1.62	0.66	0.63	0.63	0.63	0.61	0.45	0.015	0.014	0.018	0.035	0.099
BG	0.025	0.86	− 1.29	− 1.31	− 1.31	− 1.32	− 1.64	0.036	0.047	0.045	0.047	0.13