Observed differences in the levels of DDT were
examined using analysis of variance (ANOVA). Differences
were examined with a two-factor ANOVA with the
factor "station" and "period"
fixed and with the factor "period" nested
within "station"(10)) . Fixed factors
are those which are concerned with specific identifiable
treatments (eg specified areas or specified times
such as months or seasons), whereas random factors
are those concerned with a general problem (eg
increases in a broad area with samples taken from
randomly chosen locations from within that area).
The construction of hypotheses which require analyses
using a fixed factor model infers the adoption
of a more precautionary approach(11)) . A fixed
factor model will have much lower critical F values
than a random factor model and is more likely
to detect a significant difference.
Interactions
among period and area were only investigated further
if both of the main factors (Period and Station)
were significant. Significant main effects were
examined using the Student-Newman-Keuls (SNK)
multiple comparison procedure to isolate which
treatments differ from others. The approach followed
here is an internationally recommended technique
for use in environmental monitoring(12)) . For
all of the analysis of variance techniques performed
during the monitoring programme, initial analyses
were performed to ensure that the data were compliant
with the specific assumptions of analysis of variance.
These assumptions state:
-
the
data within and among samples must be independent
of each other;
-
the
variance within samples must be equal (tested
through the use of the Levene's median test);
and,
-
the
data among the samples must be normally distributed
(tested through the use of the Kolgomorov-Smirnov
test).
As
is typical with data gathered during environmental
surveys, much of the data failed to agree with
the latter two assumptions. When this was the
case, appropriate transformations such as the
logarithmic transformation were used (for DDT
Log10 transformation was used) and the data re-tested
for compliance with the assumptions.
Data
obtained from the ten sampling stations were analysed
statistically for significant differences in sediment
DDT concentrations both spatially and temporally.
A complete list of the raw data is provided in
Annex A. Summary
information of the statistical analysis conducted
and results are shown in Table 2.3b.
Table
2.3b Statistical Analysis of Total DDT in
the Sediments1
|
Source |
Degree
of freedom
|
Mean
Square
|
F
Statistic
|
P
|
Sampling
Station |
9
|
2.99
|
17.19
|
**
|
Sampling
Period
|
3
|
0.31
|
1.78
|
NS
|
Residual
|
160
|
0.17
|
|
|
SNK
result P < 0.05
VS3 WS1 DS3 SS3 ES2 TS2 NS4 PS6 MS15 MS5
Results
of the Two-Factor ANOVA followed by the SNK multiple
comparison procedure test showed that Stations
VS3, WS1 and DS3 had significantly (P < 0.05)
higher Total DDT concentrations than all other
stations except SS3. There was a degree of overlap
between stations SS3, ES2, TS2, NS4 and PS6. The
lowest concentrations of DDT were reported from
the two stations in Mirs Bay, MS15 and MS5. As
the sample sizes were small, power analysis was
conducted on the log-transformed data. Results
showed that power was high (1.00) and there was
no chance of committing a Type II error (failing
to detect a significant difference). None of the
stations showed significant temporal changes in
DDT concentration (Table 2.3b).
|