OF PSEUDOCOCCIDAE & ERIOCOCCIDAE 81 



22-29. Species 1-9 = Planococcus group of genera ; species 10-15 — Pseudococcus 

 group ; species 16 and 17 = Saccharicoccus group ; species 18 = Odococcus group ; 

 species 19 and 20 = Ceroputo group, and species 21 = Nairobia group. 



The 138 characters taken into consideration are the same as those listed in 

 table I (see Discussion, p. 31). Their different state conditions were expressed in 

 numerical code ranging from 1 to 5, but in most cases only 2-3 conditions were 

 used. The better the structure is developed the higher the value given, e.g. the 

 absence of a certain ridge was given 1, its weak development 2 and its strong 

 development 3 ; if the absence of setae or pores on a particular part of the body was 

 given the value of 1, the presence of 1-3 setae was given 2, and 4-6 for example 

 will be 3, and 6-10 will be 4 ; the small size (e.g. body) was given 1, intermediate 2, 

 and large 3. These conditions were punched on to the cards for processing on the 

 IBM 7090 Fortran IV electronic computer available at Imperial College. 



In the Principal Component method, the computer generates covariance matrix 

 but prints only the diagonal values, the sum of which (Trace) represents the amount 

 of total variance (Text-fig. I). Example of Pseudococcidae above. 



The BIGMAT calculates any desired number of latent roots (eigenvalues) and the 

 corresponding latent vectors (eigenvectors) ; as usual 10 were calculated, but only 

 the first three were used for plotting (Text-fig. II). 



Each latent root represents an axis which is perpendicular to all the other axes, 

 and the latent vectors represent the co-ordinates on these axes, thus locating the 

 species in the N-multiple superspace. The value of the latent root, expressed as a 

 percentage of the sum of diagonal values, gives the amount of variance (Trace) 

 accounted by each root. This amount is the highest for the first root and gradually 

 decreasing in the other roots. The first three in this case (29-41%, 15-96% and 

 11-33%, respectively) account for 56-7% of total variance, and these were used 

 for plotting (the remaining 7 roots account for 6-71%, 6-12%, 5-27%, 3-89%, 

 3'7°%» 3'4°% an< i 2-87%, respectively). The values of the co-ordinates are 

 calculated by multiplying the eigenvalues by the square root of the corresponding 

 latent root of each species. For the purpose of plotting an integral number was 

 added to the latent vectors of each latent root to eliminate negative values, and in 

 this case 6, 5 and 6, respectively were added to the three roots. Table 1 shows the 

 data for the pseudococcid species prepared for plotting. 



The 3-dimensional block diagrams of the first three vectors were made separately 

 for (1) all forms of Pseudococcidae and Eriococcidae studied, (2) Pseudococcidae 

 and Eriococcidae excluding the apterous males, and (3) Pseudococcidae only 

 (Figs. 1, 2 and 3, respectively). In these diagrams the space was divided by 

 horizontal planes. The first vector (axis) was represented by one side of the square 

 base (horizontal, I in Figures), the third by the other side (III, oblique in the 

 Figures), and the second vector by the vertical axis (II). For the purpose of easier 

 reference to the actual location of the species, each side of the base (and other planes) 

 was divided into 3 sections marked A, B, C for axis I, and a, b, c for axis III. The 

 resulting 9 squares are defined by the reference to the appropriate sections on the 



