By Cindy Orser | Chief Science Officer
There are two major considerations in determining a sampling strategy for a cannabis harvest or production batch: 1) the number of samples to be taken from a defined unit test batch, harvest or production; and, 2) the amount (mass) of each of those samples and whether they are tested separately or combined into one composite sample.
The State of Nevada originally addressed this statistical uncertainty by focusing on the second aspect by defining the test batch as 5 lbs. for flower with a random sample mass of 12 g or less which is roughly 0.5% sample size of lot. In the revised regulations, it would appear that this sampling guideline has been removed leaving it to each cannabis testing lab to determine the sample size. Plugging the original sampling guideline into a generic sample size calculator,[1] assuming a confidence interval (margin of error) of 15%, which means that the results would fall within +15% and -15% of the entire relevant population that was sampled with a confidence level of 95%, five (5) separate random samples would need to be taken to make up that 12 grams or less (based on the original regulations.)
As a relevant point of comparison, Colorado recently underwent a MED public rule changing exercise where the following medical cannabis sampling guidelines were introduced focusing on the first tenet or the number of samples to be taken per unit batch and not the mass of the sample in relation to the harvest or production batch (see the following below):
“At a minimum, each Test Batch of Medical Marijuana or Medical Marijuana Concentrate must be comprised of at least the following number of separately taken Samples:
- For Test Batches comprised of Harvest Batches or Production Batches weighing up to 10 pounds, eight separate Samples must be taken and combined into one Test Batch..
- For Test Batches comprised of Harvest Batches or Production Batches weighing more than 10 pounds but less than 20 pounds, 12 separate Samples must be taken and combined into one Test Batch..
- For Test Batches comprised of Harvest Batches or Production Batches weighing 20 pounds or more but less than 30 pounds, 15 separate Samples must be taken and combined into one Test Batch..
- For Test Batches comprised of Harvest Batches or Production Batches weighing 30 pound or more but less than 40 pounds, 18 separate Samples must be taken and combined into one Test Batch..
- For Test Batches comprised of Harvest Batches or Production Batches weighing 40 pounds or more but less than 100 pounds, 23 separate Samples must be taken and combined into one Test Batch..
- For Test Batches comprised of Harvest Batches or Production Batches weighing 100 pounds or more, 29 separate Samples must be taken and combined into one Test Batch.. “
Plugging Colorado’s sampling guidelines of eight (8) separate samples from a 10 lbs. batch into a sample size calculator,[2] provides for a confidence interval (margin of error) of 15%, which means that the results would fall within +15% and -15% of the entire relevant population that was sampled with a confidence level of 95%, the same parameters that were used above to determine sampling number for Nevada.
While in general, the larger your sample size (in number), the more certain you can be that the answers truly reflect the population. This indicates that for a given confidence level, the larger your sample size, the smaller your confidence interval. However, the relationship is not linear (i.e., doubling the sample size does not halve the confidence interval) and a per sample mass should have some relation to the product (i.e., a dose or a manufactured unit).
Therefore, given the pending revised cannabis sampling guidelines in Nevada with no mass or sampling parameter guidelines, we propose a combination of Nevada’s original sampling guidelines; I.e. 12 g or less per 5 lbs., would translate into selecting four random subsamples made per batch which are then combined into one composited sample for testing in the amount required to complete the composite tests based on product category. In addition, we recommend each testing lab also select a QA/QC sample at the rate of 1 per 10 samples in equivalent mass for a confirmation of findings.
[1] http://www.surveysystem.com/sscalc.htm
[2] IBID