Table 1 Associations between voice pitch features and major depressive disorder.

From: Unraveling the associations between voice pitch and major depressive disorder: a multisite genetic study

Features

Statistical functionals

CONVERGE

Replication

Combined

Beta

P

P_FDR

Beta

P

P_FDR

Power

Beta

P

P_FDR

ΔF0_iqr1-3

Interquartile range (3rd–1st)

–1.07

3.5E–51

1.1E–49

–1.09

1.46E–07

0

1

–1.07

3.4E–59

6.7E–58

ΔF0_percentile99.0

Maximum (99th percentile)

–0.97

1.2E–43

1.8E–42

–0.97

5.35E–06

0

1

–0.97

1.8E–49

1.8E–48

F0_upleveltime90

Time with F0>90th percentile

–0.80

1.5E–41

1.5E–40

–0.78

5.37E–05

0.0001

1

–0.80

1.9E–45

1.3E–44

ΔF0_kurtosis

Kurtosis

0.87

1.5E–39

1.1E–38

0.83

2.69E–04

0.0005

1

0.86

6.4E–42

3.2E–41

ΔF0_rqmean

Root quadratic mean

–0.69

6.3E–31

3.8E–30

–0.76

2.99E–04

0.0005

0.99

–0.70

6.3E–34

2.1E–33

ΔF0_quartile3

3rd quartile

–0.78

5.4E–30

2.7E–29

–0.88

1.16E–05

0

0.98

–0.80

4.1E–35

1.6E–34

ΔF0_minPos

Position of the minimum

–0.60

3.3E–25

1.4E–24

–0.57

7.05E–07

0

0.96

–0.59

7.2E–31

2.0E–30

ΔF0_amean

Mean

0.43

5.8E–20

2.0E–19

0.56

5.85E–08

0

0.88

0.45

3.0E–26

7.4E–26

ΔF0_linregc1

Slope of linear regression

–0.39

6.0E–20

2.0E–19

–0.55

3.12E–03

0.0042

0.88

–0.41

8.7E–21

1.5E–20

F0_range

Range

0.57

1.0E–19

3.1E–19

0.48

2.35E–03

0.0034

0.88

0.55

3.7E–22

8.2E–22

ΔF0_maxPos

Position of the maximum

–0.55

1.8E–19

4.9E–19

–0.61

1.36E–03

0.0023

0.87

–0.56

5.4E–22

1.1E–21

ΔF0_qregc1

1st quadratic regression coefficient

0.38

1.0E–17

2.6E–17

0.47

1.04E–04

0.0002

0.83

0.40

9.2E–22

1.7E–21

ΔF0_flatness

Flatness

–0.43

5.5E–14

1.3E–13

–0.39

2.85E–05

0.0001

0.68

–0.42

6.6E–18

1.0E–17

F0_lpc2

2nd linear prediction coding coefficient

0.33

1.3E–09

2.7E–09

0.41

2.22E–05

0.0001

0.44

0.34

6.0E–13

8.6E–13

F0_lpc4

4th linear prediction coding coefficient

–0.28

4.9E–09

9.8E–09

–0.25

1.07E–01

0.1259

0.4

–0.27

2.7E–09

3.6E–09

F0_kurtosis

Kurtosis

0.23

1.1E–05

2.0E–05

0.30

1.81E–03

0.0028

0.19

0.24

3.6E–08

4.5E–08

F0_lpc0

0th linear prediction coding coefficient

–0.29

9.0E–05

1.6E–04

–0.39

1.97E–02

0.0246

0.14

–0.30

6.1E–06

7.2E–06

ΔF0_risetime

Time with which ΔF0 is rising

–0.15

4.1E–04

6.9E–04

–0.25

1.39E–01

0.1539

0.1

–0.16

8.9E–05

9.9E–05

F0_ff0_maxSegLen

Maximum length of voiced segments with F0 > 0

0.19

6.3E–03

1.0E–02

0.12

2.69E–01

0.2837

0.05

0.19

3.2E–03

3.4E–03

F0_rqmean

Root quadratic mean

0.14

9.0E–03

1.3E–02

0.00

9.97E–01

0.9974

0.05

0.12

1.7E–02

1.7E–02

  1. The table shows the names of the prosodic phenotypes, explained in Table S2. Beta values (Beta) P values (P) and FDR corrected (FDR) are from the logistic regression analysis. Beta coefficients are derived from analyses of normalized voice features (with standard deviation of 1).Results from the CONVERGE study and an independently collected replication are shown. The column headed ‘Power’ shows the power of the replication study to detect the effect found in the discovery sample. The last three columns (‘Combined’) are results from a meta-analysis of CONVERGE and the Replication sample.
Back to article page