Biomedical Engineering Reference
In-Depth Information
Fig. 8.7
Stress plots. The adult data set using (
8.15
)inplot
A
and using (
8.16
)inplot
B
.The
children data set using (
8.15
)inplot
C
and using (
8.16
)inplot
D
Fig. 8.8
Shepard plots for
n
i
=
3 with adult data set (plot
A
) and with children data set (plot
B
)
a low-dimension MDS plot. For children, in Fig.
8.7
-C we see that for (
8.15
), the
stress plot “knee” is at
n
i
=
3, while in Fig.
8.7
-D for (
8.16
) the stress plot “knee” is
at
n
i
=
4. In this case, the opposite situation arises: that (
8.16
) is more demanding.
For the sake of uniformity, we shall choose
n
i
=
3 for all MDS plots.
The Shepard plots are depicted in Figs.
8.8
-A and
8.8
-B for (
8.15
) and (
8.16
),
in adults and children, respectively. A Shepard plot is considered valid when the
dots follow the 45 degree line. For example, if the dots present a low dispersion but
following a curve (instead of the 45 degree line), it suggests that some nonlinear
relation exists. Usually, the Shepard plots are better the higher the dimension, but
“stabilize” for the
n
i
value obtained from the stress plot. That is why the stress
and the Shepard plots are usually “redundant”. From both figures we conclude that
(
8.16
) delivers better dispersion than (
8.15
).
With the impedance data at hand, one can now assess the information from the
MDS plots for
n
i
=
3. The mapping obtained with the MDS for relations (
8.15
)are
given in Figs.
8.9
-
8.10
for the comparison between healthy adults and those with
COPD and KS; respectively in Figs.
8.11
-
8.12
for the comparison between healthy
