Skip to main content
. 2014 Sep 18;15(6):883–896. doi: 10.1007/s10162-014-0487-3

TABLE 1.

The parameters of the linear regression model fit to the data in Figure 5 (on a log-log scale, see Data analysis).

Linear regression model
0.3 < f probe < 16 kHz 3 < f probe < 16 kHz
CAP-STC slope 0.27 (p < 0.001) 0.35 (p < 0.001)
Slope difference SF-GD vs CAP-STC 0.11 (p = 0.066) n.s.
SF-STC vs CAP-STC 0.59 (p < 0.001) n.s.
CAP-STC intercept −0.40 (p = 0.007) −0.74 (p = 0.009)
Intercept difference SF-GD vs CAP-STC −0.33 (p = 0.12) 0.11 (p = 0.001)
SF-STC vs CAP-STC −0.13 (p < 0.001)

The model was fit to data over either the full range of probe frequencies or to data obtained only for high frequencies (>3 kHz). Both models were significant (p < 0.001) and explained 82.7 and 44.0 % of variability in sharpness of tuning, respectively (based on the adjusted R 2). In the first “full” model, most of the variance was predicted by the probe frequency (R 2 = 0.42) while in the other “restricted” model by the variables coding the test type (R 2 = 0.26). The intercept estimate for SF-STC in the first model was omitted due to significant change in slope (i.e. significant interaction between test type and frequency).