my.lda=lda(class ~.,x)

above code, my.lda is a LDA object and is composed of eight

nts, namely $prior, $counts, $means, $scaling, $lev,

N, $call, $terms and $xlevels. Among them, $prior,

s, $means and $scaling are the main components for further

e 3.3 shows two data sets with two clusters but different

ons of data points from two clusters. Figure 3.3(a) shows a data

an even data distribution across two classes but Figure 3.3(b)

data set with an odd data distribution across two classes. Two

ad different centres, but had the same covariance matrix shown

ܵௐൌቀ⁴

2

2

4^ቁ

(a) (b)

wo data sets for illustrating how LDA and the Bayes rule work for data sets with

d data point number. (a) Both classes (the triangles and the crosses) were

of 200 data points. (b) The triangle class was composed of 200 data points, but

ass was composed of 2,000 data points.

e 3.4 shows the densities of the LDA predictions (ݕො஺ and ݕො஻) for

sets shown in Figure 3.3. Because two clusters had an even data

on across two classes (both have 200 data points) in the data set

n Figure 3.3(a), two a priori probabilities were identical, i.e.,

ൌߨ஼௥௢௦௦ൌ0.5. The two density functions of two classes