journal menu
![[Figure 4]](gm5051fig4mag.jpg)
|
|
Figure 4
Properties of the ICA analysis for thaumatin. (a, b, c) Three plots of contributions from 22 thaumatin data sets for three ICA components. Dose increases monotonically with the data set number. (a) Contributions to the first component indicate that the first component resembles well the (dρ/dD) estimator for initial changes. (b) Contributions from each data set to the second component of ICA. The monotonically rising pattern of contributions in combination with significantly higher absolute values for data sets acquired at higher doses indicates that the 2mFo − DFc-type map generated by the second component corresponds to the dose being ∼1.5–2 times higher than the one achieved in the experiment. (c) Contributions from thaumatin data sets to the third component of ICA. The parabolic shape of the dependence indicates a large contribution from the (d2ρ/dD2) (non-linear) estimator. According to that shape, the rate of specific radiation changes with dose imaged by the third component corresponds to the combination of rate changes (d2ρ/dD2) at higher doses together with initial deceleration (dρ/dD) at higher doses. (d) Scatter plots of ICA results for thaumatin with kurtosis (right) and skewness (left) as the ICA-optimized targets. The ICA results are projected onto the plane defined by the first and the second ICA components. The units are arbitrary because the method is scale-independent. However, their ratio carries information about the relative magnitude of components. The colour gradient represents the density of the points. Both targets show a slightly asymmetrical distribution along both axes, which indicates that full statistical independence between the ICA components was not achieved. |
![[Figure 4]](gm5051fig4.jpg)
 | JOURNAL OF SYNCHROTRON RADIATION |
ISSN: 1600-5775
