Analyte   Adj. R2      F          p-value   # Coefficients   # Sig.
CBCA      0.075     1.260    0.316       9                      N/A
CBDA      0.246     2.052    0.087       9                       N/A
CBNA      0.526    4.573     0.002       9                          3

Phase 2 analysis (trimmed data)

For phase 2 of the analysis, the <LOD substitution values and all values that exceeded the 95% confidence interval upper limit computed in the initial investigation were removed. The rationale here was to determine if data would exhibit a more normalized distribution once upper outliers and <LOQ values, which may have contributed to the observed right-tailed skew, were deleted.
While oil density retained its normal distribution, the normality of distribution for many of the analytes was noticeably improved from phase 1 (see Table 2). Histograms and Q-Q plots were constructed for oil density, and each cannabinoid in the phase 2 data set, as shown in the supplemental data addendum. The higher number of statistically normally distributed analytes allowed for an increased number of regression tests in phase 2.
The supplier and cultivar were coded into new dichotomous variables in phase 2 using the same organization as conducted in phase 1 to investigate the possible correlation between cannabinoid content variation and suppliers and cultivars. Not all suppliers or cultivars were included in the phase 2 regression analysis because some samples were excluded as a result of trimming data at the 95% CI upper limit.
Linear regression analysis was conducted using the phase 2 (trimmed) and coded supplier data. The supplemental data addendum shows the regression model summaries, ANOVA tables, and coefficient tables for the phase 2 supplier regression tests. Many of the regression models for supplier were significant and had a higher number of significant coefficients than in phase 1. A summary of the supplier regression analysis with suppliers is provided in Table 5.
Table 5: Phase 2 regression of analytes and suppliers