Results and Discussion
Statistical tests for normal distribution were conducted before any
regression analysis. The Shapiro-Wilk method tests the null hypothesis
that the data are normally distributed, and was selected due to the
small number of samples, particularly after removing upper outlier
values in the phase 2 data set. Normality tests are essential since many
other statistical tests presume a normal distribution of data, notably
the dependent variables in linear regression. Table 2 summarizes the
results of the Shapiro-Wilk normality tests for both raw and trimmed
data.
In the phase 1 data set, only density and three of the fifteen (20%)
cannabinoids (CBCA, CBDA, CBNA) had a statistically normal distribution
(p > 0.05). Conversely, the distribution was statistically
normal for nine of fifteen (60%) of the analytes in the phase 2 data
set. CBCA, CBC, CBD, CBL, CBN, and THC were not normally distributed in
the phase 2 data set. The Shapiro-Wilk test showed increases in
significance (p -values) for individual analytes in the phase 2
data set, but it was not enough to pass the normality test at 0.05
significance.
Table
2 (a-b): Shapiro-Wilk tests for normal distribution