0851996884.pdf

From these data, the following descriptive

statistics can be computed: (i) for the

females, n 1 = 30, x– 1 = 15.83 days, σ 12 = 31.18;

(ii) for the males,n 2 = 30, x– 2 = 9.80 days, σ 22 =

19.41; and the test is thus:

which is greater than 1.96. So there is a sig-

nificant difference between male and female

adults longevity, at a probability level of 5%.

Figure 20.6 shows the graph summarising

this difference.

Conclusions

QC workers usually collect a lot of quantita-

tive and qualitative data. These data have to

be synthesized properly to assess the quality

of beneficial organisms as biocontrol agents.

For this, statistical tools provide substantial

help, both in summarizing the main features

of the samples collected and in checking

whether the mass-reared animals are satisfy-

ing predefined criteria. The statistical proce-

dures discussed in this chapter are those

which only consider samples described by

one or two variables. Further information

about these tests can be found in statistical

handbooks. There are a number of software

packages that include the above-described

tests, such as: (i) SAS®(SAS Institute, Inc.

(1990); http://www.sas.com/products/

sassystem/index.html); (ii) STATISTICA®

(http://www.statsoftinc.com/); (iii) SPSS®

(http://www.spss.com/); and (iv) SYSTAT®

(http://www.spssscience. com/systat/). Of

course, studying the quality of a mass-pro-

duced animal cannot be carried out by

describing only two traits. So, in most of the

cases, more than two traits are measured

simultaneously. For example, in order to

assess the quality of the aphid parasitoid

Aphelinus abdominalis, at least five biological

traits have to be quantified: (i) emergence

rate; (ii) sex ratio; (iii) adult size; (iv) female

fecundity; and (v) adult mortality. In order to

handle these kinds of data sets, the proce-

dures described here should be generalized

using multivariate (also called multidimen-

sional) methods. More accurately, multidi-

mensional correlations can be computed and

tested in order to see whether there is some

redundancy between the different traits

studied. Besides this, multidimensional con-

fidence intervals can also be constructed in

order to check whether the averages of the

different traits quantified satisfy the prede-

fined multidimensional standard. However,

the corresponding methods, which are just a

generalization of those presented here, can-

not be used without some specific statistical

computer software package running on

microcomputers or on mainframes, and

these are not described here. Finally, the only

way to prove that a mass-produced biocon-

trol agent is of a good quality is to check,

after it has been released in the field,

whether its pest-control efficiency is suffi-

cient. Therefore, there is also a need for sta-

tistical methods to determine the correlation

between such a field efficiency and the sim-

ple traits that are quantified in the labora-

tory. As the corresponding methods are also

based on multivariate approaches, special

Statistical Methods for Quality Control 313

Longevity of males in days:

13 11 15 7 9 15 9 11 14 7 9 12 8 16 16

171068616102131160935

t =−1583 980 3118 30 1941 30 464.././ ./ ., + =

20

15

10

5

Females Males

Longevity (days)

Fig. 20.6.Average (±SE) longevity of females and

males of T.brassicae.