Information Technology Reference
In-Depth Information
The general schema of performing such a test is given by the following
procedure:
1. The measure of a statistical parameter
Q
in two populations (for example the
mean blood pressure);
2. The determination of the probabilistic law
F
which the chosen parameter follows
(
F
gives, for each possible value of the parameter
Q
, the probability of having
samples where the parameter assumes that value);
3. The calculation of the value
V
of the chosen parameter, in the population where
the cause under investigation is acting;
4. Definition of the confidence interval of the test, that is, the level of error proba-
bility
E
which is tolerated in the choice of the hypothesis (for example, 3% or
5%);
5. The localization of the observed value
V
in the distribution probability
F
;
6. The evaluation of the probability
P
V
which
F
assigns to the event
{
Q
≥
V
}
(or
, in the case of a symmetric distribution);
7. The rejection of the null hypothesis if
P
V
≤
{|
Q
|≥|
V
|}
E
, with the confidence
E
, or accep-
tance in the other case (with confidence
E
).
In the previous procedure,
P
V
is also called the
p-value
of
V
, while the value
V
0
such that the probability of
{
Q
≥
V
0
}
is less than
E
, is called the
critical value
for
Fig. 7.5
Statistical inferential value of test parameters