Geoscience Reference
In-Depth Information
that exist within the data set, we will now pretend that we do not know the
relative contributions that the three source rocks made to the thirty sediment
samples. Instead, we have a record of the mineral contents of thirty sediment
samples stored in the i le
sediment_1.txt
.
clear
x = load('sediments_1.txt');
h e aim of the PCA is now to decipher the statistically independent
contribution of the three source rocks to the sediment compositions. We can
display the histograms of the data and see that they are not perfectly Gaussian
distributed, which means that we cannot expect a perfect unmixing result.
subplot(1,3,1), histogram(x(:,1))
subplot(1,3,2), histogram(x(:,2))
subplot(1,3,3), histogram(x(:,3))
We display the proportions of the three minerals in the thirty samples along
the sedimentary section. In this graphic we can see weak correlations and
anti-correlations between the proportions of the three minerals.
plot(x(:,1:3)), grid
legend('Min1','Min2','Min3')
xlabel('Sample ID')
ylabel('Quantity')
Before running the PCA we dei ne labels for the various graphics created
during the exercise. We number the samples 1 to 30, with the minerals being
identii ed by four-character abbreviations
Min1
,
Min2
and
Min3
.
for i = 1 : 30
samples(i,:) = [sprintf('%02.0f',i)];
end
minerals = ['Min1';'Min2';'Min3'];
We can explore the correlations between the minerals in pairwise bivariate
scatter plots. We observe a strong negative correlation between the i rst
and second mineral, a weak positive correlation between the i rst and third
mineral, and a moderate negative correlation between the second and third
mineral.
subplot(1,3,1),
plot(x(:,1),x(:,2),'o')
xlabel('Mineral 1')
ylabel('Mineral 2')
subplot(1,3,2),
plot(x(:,1),x(:,3),'o')
