238
M. Gevrey . S. Lek' T.Oberdorff
aligned on the diagonal of the perfeet-fit line. (Fig. 12.2a). There are many low
values of TSR, whieh are underestimated, but there is a eomplete independenee
between the residuals values and the estimated values (n=136, r= 0.045, P=0.606)
(Fig. 12.2b). Nevertheless, there is an exaggerated clustering of residuals at zero in
order for the distribution to be normal. The assumption of normality may not hold.
Lillieffors test of normality of residuals gives a maximum differenee of 0.225
(P 500
300 (b)
'"
(a)
co
200
::s 400
CIl
co
~ 100
. .
> 300
. .
"0
~_. ,
co
"d
0
-,p. ':"
~
...
200
CIl
e
~ -100
~
100
'"
-200
r.:.l
-300
100 200 300 400 500
0
200
400
600
Observed va lues
Estimated values
60 (c)
50
40
30
20
10
0
0
0
0 0 0 0
0
0
0
'<;' ""r ~
N
..... \0 oe 0
Residuals
Fig. 12.2. (a) Relationship between observed and estimated values of TSR,
(b) Relationship between the residuals and the estimated values of TSR,
(e) Distribution of residuals (observed values- estimated values of TSR).
12.4.2
Sensitivity Analysis
Here are given the results of the investigation of the two sensitivity methods.