e was the location of the crossover point between two posterior

ty curves. In Figure 3.5(a), two posterior probability curves were

n the centre (zero) because two classes of data points had the same

In Figure 3.5(b), the crossover point of two posterior probabilities

ated from zero towards the left side because the data set had an

distribution across two classes.

valuate the performance such as the discrimination power of a

ed LDA model, new data should be inputted into a constructed

del to examine whether the outputs of a constructed LDA model

pected. When inputting new data to a constructed LDA model,

uts of the LDA model are called the predictions. To make

n based on a constructed LDA model for new data, the predict

is called,

my.output=predict(my.lda,newdata,···)

main components of my.output include $class, $posterior

$x is composed of the predictions ሺݕො) which are the mapped data

timised projection direction (ܟෝ) from the independent variables.

erior probabilities are saved in a matrix of two columns for two

nd is named as $posterior. Its first column is composed of

rior probabilities for the first class and the second column is

d of the posterior probabilities for the second class of the data set

h the LDA model is constructed. If the first posterior probability

is greater than the second posterior probability in the same row,

point at that row is assigned a label of the first class. Otherwise,

ned a label of the second class. Based on a default threshold (0.5)

the posterior probabilities for decision-making, the posterior

ty $posterior is converted to the predicted binary variable

binary and is named as $class in the output object

tput) of the predict function. For a row, if the second

probability is greater than the first posterior probability, $class

therwise, it is zero.