. 2015 Feb 26;39(3):195–214. doi: 10.1007/s10919-015-0209-5

Table 2.

Results of the linear mixed models on the acoustic structure of vowels

Parameter	Vowel	Emotion^a		Encoding condition^a		Gender^a
Parameter	Vowel	$χ_{(3)}^{2}$	p ^b	$χ_{(2)}^{2}$	p ^b	$χ_{(1)}^{2}$	p ^b	Estimates ± SE^c
F0	a	15.34	.005**	0.42	1	43.56	<.001***	70.49 ± 9.72
	e	15.02	.005**	0.86	1	45.07	<.001***	69.72 ± 9.30
	i	14.48	.007**	3.41	.540	32.12	<.001***	70.18 ± 11.46
HNR	a	1.35	1	1.28	1	21.15	<.001***	0.046 ± 0.01
	e	3.32	1	0.19	1	22.99	<.001***	0.045 ± 0.009
	i	11.91	.023*	2.59	.822	41.78	<.001***	0.076 ± 0.010
Amplitude ratio	a	5.61	.397	52.07	<.001***	5.09	.072	−0.221 ± 0.099^d
	e	1.74	1	26.15	<.001***	27.61	<.001***	−0.651 ± 0.117^d
	i	11.65	.026*	3.79	.450	32.44	<.001***	−1.028 ± 0.164^d
Peak frequency	a	21.42	<.001***	24.22	<.001***	0.76	1	0.067 ± 0.0774^d
	e	8.13	.13	17.53	<.001***	6.99	.024*	0.169 ± 0.064^d
	i	0.55	1	24.58	<.001***	6.50	.032*	0.121 ± 0.048^d
First formant	a	4.66	.198	19.08	<.001***	5.00	.025*	0.190 ± 0.086^d

Asterisks mark the significance level at *p < .05; **p < .01; ***p < .001

^aStatistical values are obtained from the model comparison (full model to reduced model excluding the respective predictor)

^b p value adjustments (Bonferroni correction) were done for the different vowels within one acoustic parameter and one predictor

^cEstimates for the predictor gender were gained from the LMM, with male speakers included in the intercept. Estimates refer to the female speakers in comparison to the male speakers

^dValues base on log transformed data