1 00:00:00,000 --> 00:00:08,100 Welcome to the Cannabis Data Science Meetup Group. 2 00:00:08,100 --> 00:00:16,900 I wanted to share with you some of the research and how eager cannabis data scientists like yourself can help out. 3 00:00:16,900 --> 00:00:21,900 I'll go ahead and sort of jump into the material that I wanted to talk about. 4 00:00:21,900 --> 00:00:27,700 But before I do that, why don't I give all of you a chance to talk about what you may want to talk about? 5 00:00:27,700 --> 00:00:31,400 In no particular order, maybe starting in my top corner. 6 00:00:31,400 --> 00:00:34,100 Rick, welcome to the group. 7 00:00:34,100 --> 00:00:44,400 I'd be curious to hear about what you'd like to get out of cannabis data and maybe some interesting projects you're working on and things that you would like to see move forward. 8 00:00:44,400 --> 00:00:48,100 Sure. So, hey everyone, my name is Rick. 9 00:00:48,100 --> 00:00:54,200 I have a background working professionally in big data in the healthcare industry. 10 00:00:54,200 --> 00:01:02,400 So, and I'm also a long time lover of cannabis, whose state has been recreational here for a few years. 11 00:01:02,400 --> 00:01:09,500 So, I've taken advantage of that and began cultivating and most of the passion has gone into breeding. 12 00:01:09,500 --> 00:01:19,400 That is my primary interest in data is compiling as much as I can and creating relations for strains and effects. 13 00:01:19,400 --> 00:01:28,800 And I started looking into it, saw what you guys were doing and it was, you know, right on the same page as what I'm looking into right now. 14 00:01:28,800 --> 00:01:39,300 Ultimately, I'd like to build an app that is geared towards breeders and facility owners that allow them to use their data more efficiently. 15 00:01:39,300 --> 00:01:47,500 I work with some of the facilities in Michigan here and see just all the untapped potential and data going out the window. 16 00:01:47,500 --> 00:01:51,200 So, I'm working on something to fix that here locally. 17 00:01:51,200 --> 00:01:55,300 Simply phenomenal, Rick. Since you weren't here last week, I'll go ahead and recap. 18 00:01:55,300 --> 00:01:59,100 All of the varieties are of quite interest, right? 19 00:01:59,100 --> 00:02:04,900 So, go to any retailer and they're going to be selling by variety. 20 00:02:04,900 --> 00:02:08,400 So, it's a long time tradition in the cannabis industry. 21 00:02:08,400 --> 00:02:17,300 And what we like to do is marry sort of folklore and things that people use every day to statistics. 22 00:02:17,300 --> 00:02:23,800 And so, if people are using these strain names every day, then we're going to see if there's any rhyme or reason to it. 23 00:02:23,800 --> 00:02:29,500 And it seems that, of course, underlying this is the chemical variability. 24 00:02:29,500 --> 00:02:32,300 And where does the chemical variability come from? 25 00:02:32,300 --> 00:02:35,000 Well, that may come from growing conditions, right? 26 00:02:35,000 --> 00:02:39,200 It may come from genetics, could come from probably both. 27 00:02:39,200 --> 00:02:44,200 Love that you're trying to understand this and then you're following it through, right? 28 00:02:44,200 --> 00:02:47,900 From seed to effect, right? 29 00:02:47,900 --> 00:02:57,400 Because, right, not only do you go seed to sale, but what happens after sale, people ingest this and presumably have some sort of effects. 30 00:02:57,400 --> 00:03:01,500 And we're trying to determine what, if any, are those effects. 31 00:03:01,500 --> 00:03:11,900 And we kind of hit on last week that at least I think the appropriate way to go about this is just having just a null hypothesis that, 32 00:03:11,900 --> 00:03:16,000 OK, let's just say that maybe cannabis has no effect on people. 33 00:03:16,000 --> 00:03:18,800 Of course, there's lots of evidence to the contrary. 34 00:03:18,800 --> 00:03:26,400 So, you know, that may be a easily falsifiable hypothesis, but that's a good starting point. 35 00:03:26,400 --> 00:03:36,600 So then, right, you'll want to one by one see what effects these various compounds people are ingesting are having. 36 00:03:36,600 --> 00:03:40,900 At least that's the way I think a nice scientific way to go about this would be. 37 00:03:40,900 --> 00:03:45,300 It's brilliant. You're right on the money here. 38 00:03:45,300 --> 00:03:48,400 This is what people are working on. 39 00:03:48,400 --> 00:03:57,200 And I think you actually have a pretty sophisticated knowledge of this because, as I'll point out to you today, 40 00:03:57,200 --> 00:04:09,800 just say taking into consideration that there are different varieties of cannabis is actually not apparent to many. 41 00:04:09,800 --> 00:04:14,500 And may not be included in the conversation and that has implications. 42 00:04:14,500 --> 00:04:17,800 But I'll point some of that out later today. But enough of me rambling. 43 00:04:17,800 --> 00:04:26,100 Isaac, good to see you today. Curious if you have any projects that you want to see moved forward here in the coming weeks. 44 00:04:26,100 --> 00:04:30,500 Good to see you all again. And I'm Isaac. 45 00:04:30,500 --> 00:04:38,200 I work for MCR Labs located in Massachusetts, cannabis testing lab. 46 00:04:38,200 --> 00:04:43,700 Well, this is actually for both myself and the place I work for. 47 00:04:43,700 --> 00:04:47,100 Kind of my main focus is on lab shopping. 48 00:04:47,100 --> 00:04:55,100 I mean, for the for my employer, definitely as we're trying to do proper science, 49 00:04:55,100 --> 00:05:00,400 we're losing business to those who are fudging numbers. 50 00:05:00,400 --> 00:05:15,500 And for myself, I mean, as a cannabis consumer, this is just MC a lot of health risk and just consumer fraud that is rampant in the industry. 51 00:05:15,500 --> 00:05:23,000 And I feel I'm in a particular position that allow me to do something about it. 52 00:05:23,000 --> 00:05:38,100 So, yeah, my main and I am actually working on fraud detection. 53 00:05:38,100 --> 00:05:41,900 And actually, I'm here today with a question. 54 00:05:41,900 --> 00:05:50,500 I'm hoping to prove to prove your brain to see your thoughts on that. 55 00:05:50,500 --> 00:05:55,000 So it's about using Benfors law. 56 00:05:55,000 --> 00:06:04,400 And I think you have that would assume you have at least heard of it is about the distribution of naturally occurring numbers. 57 00:06:04,400 --> 00:06:15,200 Contrary to intuition is it follows a kind of exponential curve that there is more ones than twos. 58 00:06:15,200 --> 00:06:27,500 And but for fraud detection in cannabis, the numbers, specifically the THC values are all clumped together around 20 percent. 59 00:06:27,500 --> 00:06:37,000 Right. And that's if you go from the textbook kind of manual, that's a no no for application of Benfors law. 60 00:06:37,000 --> 00:06:52,400 But then I think I found a method or operation that is to raise the numbers to a high power, a fourth or fifth power. 61 00:06:52,400 --> 00:07:01,600 By doing that, it seems to be able to stretch the numbers into several magnitudes into a realm where Benfors law apply. 62 00:07:01,600 --> 00:07:08,200 So the most straightforward case is take integers from, say, one to a thousand. 63 00:07:08,200 --> 00:07:13,900 But first, I won't apply it right because it's sequential numbers. 64 00:07:13,900 --> 00:07:19,100 But if you keep those numbers, then first law does appear. 65 00:07:19,100 --> 00:07:31,500 So I'm just wondering the possibility of even stretching the power of Benfors law by allowing the operation of. 66 00:07:31,500 --> 00:07:41,500 Raising the data set to a high power to artificially increase the the magnitude. 67 00:07:41,500 --> 00:07:48,200 So I'm hoping to hear your thoughts on that direction. 68 00:07:48,200 --> 00:08:00,100 And also, in addition, as my colleague, Yasha just joined, we were thinking of another reason that. 69 00:08:00,100 --> 00:08:05,400 Well, as we all know, THC values has been rising for a while. 70 00:08:05,400 --> 00:08:12,500 And the easiest explanation is, well, there could be a lot of explanations. 71 00:08:12,500 --> 00:08:20,900 LAP shopping is definitely one, but there could be also that the girls are getting better at cultivating, 72 00:08:20,900 --> 00:08:29,700 or it could be that they are simply getting better at selecting the best gene to grow. 73 00:08:29,700 --> 00:08:39,500 So we're interested in finding out how much each of those potential causes contribute to the. 74 00:08:39,500 --> 00:08:44,700 Graduate growth of THC content that we've been seeing in the Washington data should show this. 75 00:08:44,700 --> 00:08:48,200 Right. Yasha, what was your input there at the end? 76 00:08:48,200 --> 00:08:54,500 The watch the data that you guys have from Washington should should answer two of those questions, 77 00:08:54,500 --> 00:09:05,200 whether a single. Is it that genetics that yield higher THC are being used more often? 78 00:09:05,200 --> 00:09:16,100 And that leading to average of higher potency or is a single strain tested by the same grower at the same lab growing in potency or staying the same? 79 00:09:16,100 --> 00:09:24,300 I'm going to basically let you in on the golden way that you can actually go about attempting to answer this question. 80 00:09:24,300 --> 00:09:28,100 Right. You came here. So here it is. Right. 81 00:09:28,100 --> 00:09:39,400 Ultimately, the problem you're going to run into is, well, how do you identify which lab results are fraudulent and which ones are not? 82 00:09:39,400 --> 00:09:50,100 Right. And so basically at the end of the day, right, you'd want a zero or one zero, not fraudulent, one fraudulent, 83 00:09:50,100 --> 00:09:57,700 because ultimately you're going to need to predict if the lab results fraudulent or not. 84 00:09:57,700 --> 00:10:03,600 And I think, yes, exactly. You could use Bamford's law to a simple way. 85 00:10:03,600 --> 00:10:12,200 Right. If you built this logic regression, so zero or one trying to predict if it's fraudulent or not, you could say use Bamford's law. 86 00:10:12,200 --> 00:10:16,600 You could you say you could maybe maybe you know this better than I. 87 00:10:16,600 --> 00:10:26,100 I'm just kind of conjecturing here, but maybe as your regressors, you would use things like I don't know how you would incorporate this into your regressors, 88 00:10:26,100 --> 00:10:29,700 but somehow have the number of digits or something of that sort. 89 00:10:29,700 --> 00:10:33,600 Actually, in fact, I don't know how you would build out the regression. 90 00:10:33,600 --> 00:10:42,800 But the first the first and that actually may be sort of the the beauty or not the beauty, but the the novel insight you have. 91 00:10:42,800 --> 00:10:48,600 Right. Maybe if you wanted to patent some fraud detection device, you just have a really good algorithm. 92 00:10:48,600 --> 00:10:51,900 Well, first things first, you need the zeros or ones. 93 00:10:51,900 --> 00:10:57,400 And so now that I've been teasing it, well, how do you actually get about getting that data set? 94 00:10:57,400 --> 00:11:09,900 Well, it so happens that I do believe this data exists and you can get your hands on it in that we've been looking at the Washington State data. 95 00:11:09,900 --> 00:11:14,800 That's because Washington State has a pretty open Freedom of Information Act request. 96 00:11:14,800 --> 00:11:24,700 So you can get these data points specifically and what you'd be looking for or as some of you may have known, 97 00:11:24,700 --> 00:11:36,600 if you look in the news, the former laboratory in Washington State Praxis Laboratory was was called falsifying samples. 98 00:11:36,600 --> 00:11:45,900 And, you know, these lab results were entered into the traceability system, which you can then get those those results. 99 00:11:45,900 --> 00:11:52,300 That would basically be your subset of known fraudulent results. 100 00:11:52,300 --> 00:12:00,400 So that basically you would just say, OK, you know, these are the lab results that we know are fraudulent. 101 00:12:00,400 --> 00:12:05,300 One, the rest of them, you would assume are non fraudulent. 102 00:12:05,300 --> 00:12:09,800 But as you will point out, that's a big assumption, right? 103 00:12:09,800 --> 00:12:21,300 Because we don't actually know for certain t right that none of the other laboratories were not were not falsifying lab results. 104 00:12:21,300 --> 00:12:25,400 So that's sort of a hiccup in your analysis. 105 00:12:25,400 --> 00:12:32,300 But you could potentially just, I guess, look at the Praxis Laboratory lab results. 106 00:12:32,300 --> 00:12:43,700 And so then for that time window, you would just say, OK, these are all the lab results that Praxis Laboratory tested that weren't falsified. 107 00:12:43,700 --> 00:12:47,000 These are the ones that were falsified. 108 00:12:47,000 --> 00:12:55,700 Is there any way to detect to a significant degree of certainty that those were, in fact, falsified? 109 00:12:55,700 --> 00:13:00,400 And that's where your training, your prediction model would come in. 110 00:13:00,400 --> 00:13:05,800 So you say, oh, you know, and that's, I guess, yet another tricky part. 111 00:13:05,800 --> 00:13:09,100 So so that's at least the first tricky part solved. 112 00:13:09,100 --> 00:13:11,600 How do you get your training data set? 113 00:13:11,600 --> 00:13:16,700 The second tricky part is how do you actually predict? 114 00:13:16,700 --> 00:13:19,700 So that's actually going to all have to chew on that. 115 00:13:19,700 --> 00:13:21,800 But maybe you've thought about that. 116 00:13:21,800 --> 00:13:23,000 Any questions at this point? 117 00:13:23,000 --> 00:13:23,900 I'll show you. 118 00:13:23,900 --> 00:13:27,300 There's there's just multiple ways to cheat by a lab. 119 00:13:27,300 --> 00:13:33,200 There's the way that they prepped the sample, just fudging numbers. 120 00:13:33,200 --> 00:13:40,200 And there can be probably four or five different ways by which they can do so. 121 00:13:40,200 --> 00:13:43,000 So we need examples of each to be the ones. 122 00:13:43,000 --> 00:13:47,800 And then a clean data set that's just definitely clean. 123 00:13:47,800 --> 00:13:49,600 Right. And that's a good point. 124 00:13:49,600 --> 00:13:57,200 And so that would essentially be I think we kind of mentioned on this last week where there's so many variables. 125 00:13:57,200 --> 00:13:59,500 Right. And each of these are something that you would control for. 126 00:13:59,500 --> 00:14:02,700 And this is the thing about statistical models is right. 127 00:14:02,700 --> 00:14:07,300 They very laser focused in on the data at hand. 128 00:14:07,300 --> 00:14:16,100 Right. So if you train your model on data and you only had, like you said, the one way that fraud was committed, 129 00:14:16,100 --> 00:14:21,600 you know, committed, then you would miss potentially other ways. 130 00:14:21,600 --> 00:14:26,600 So maybe they are somehow preserving the digits. 131 00:14:26,600 --> 00:14:30,000 So maybe they would somehow preserve Bamford's law. 132 00:14:30,000 --> 00:14:32,000 Something else is going on. 133 00:14:32,000 --> 00:14:33,300 So that's a possibility. 134 00:14:33,300 --> 00:14:37,700 And in fact, that may just be where we just kind of need to expand the data set. 135 00:14:37,700 --> 00:14:41,800 There is the one case in Washington state where I think you could actually get the data. 136 00:14:41,800 --> 00:14:46,000 There have been other cases of fraud, like in Michigan. 137 00:14:46,000 --> 00:14:49,600 I think there was a case that may be debated. 138 00:14:49,600 --> 00:14:52,300 I think that one may be contested. 139 00:14:52,300 --> 00:14:57,700 And then I think there could have been an occurrence in Nevada and perhaps California. 140 00:14:57,700 --> 00:15:06,300 But ultimately, I think what you would want to seek are the results that you know are fraudulent 141 00:15:06,300 --> 00:15:12,300 versus the ones that you suspect aren't and see if there's any systematic difference. 142 00:15:12,300 --> 00:15:20,500 Because otherwise, and this is actually unfortunately what we've kind of seen is basically maybe not wrong, 143 00:15:20,500 --> 00:15:23,600 but I don't think it was fruitful. 144 00:15:23,600 --> 00:15:30,200 But basically, like in Washington state, right, you just see people take simple averages by lab 145 00:15:30,200 --> 00:15:37,600 and just kind of complain that a particular lab may have, you know, high on average results, 146 00:15:37,600 --> 00:15:40,700 which, you know, it's not a good look, right, 147 00:15:40,700 --> 00:15:45,900 if you're producing results that are significantly different than another laboratory. 148 00:15:45,900 --> 00:15:52,200 But that I don't think, right, there's many ways that people can go about explaining that away, right. 149 00:15:52,200 --> 00:15:55,700 There's the, oh, you know, good growers will choose our lab. 150 00:15:55,700 --> 00:16:01,000 As much as it may be like a signal of like, oh, maybe we should look here or look there. 151 00:16:01,000 --> 00:16:04,800 I don't think there is much concrete evidence. 152 00:16:04,800 --> 00:16:13,700 Because the other thing to kind of consider is there's sort of permitted ways that people can kind of push the numbers up. 153 00:16:13,700 --> 00:16:17,800 Or maybe they're not permitted, but they're definitely kind of grade lines. 154 00:16:17,800 --> 00:16:23,400 So, for example, I don't know if this is, this wouldn't really affect total THC or CBD, 155 00:16:23,400 --> 00:16:30,700 but I know there was an effort to just add as many cannabinoids as you can to your panel, right. 156 00:16:30,700 --> 00:16:39,400 Because if you're testing for CBL and CBT and CBCA and CBDV, 157 00:16:39,400 --> 00:16:47,000 so sort of just the more cannabinoids that would just push up the total cannabinoids. 158 00:16:47,000 --> 00:16:51,000 So that was kind of happening. That's kind of the case in California. 159 00:16:51,000 --> 00:16:57,000 So once again, I think there's many factors going on, but that's definitely one of them is 160 00:16:57,000 --> 00:17:04,300 there's definitely been a push to increase the panel, which isn't actually necessarily a bad thing. 161 00:17:04,300 --> 00:17:08,600 But as you pointed out, there may be other things going on. 162 00:17:08,600 --> 00:17:17,100 With that one specifically, so the more cannabinoids you have, the likely the lower THC will be 163 00:17:17,100 --> 00:17:21,900 because it'll get rid of the coelusion problem, especially with Delta-8, for example, 164 00:17:21,900 --> 00:17:23,900 that coeluses with Delta-9. 165 00:17:23,900 --> 00:17:30,100 And in a article about what happened in Washington, 166 00:17:30,100 --> 00:17:40,500 it does reference a price point at which buyers buy cannabis, which is at 20% THC. 167 00:17:40,500 --> 00:17:45,500 And I think THC by most dispensaries is defined as THC plus THCA. 168 00:17:45,500 --> 00:17:52,500 So what they think once is, and I think the number that most people look to is how much THC is in there, 169 00:17:52,500 --> 00:17:55,100 not how many total cannabinoids are in there. 170 00:17:55,100 --> 00:18:01,400 So I think what they're trying to do is specifically how can we increase THC for that final sale to be higher. 171 00:18:01,400 --> 00:18:08,100 I may need some clarification on your second point, but the first point was actually clever 172 00:18:08,100 --> 00:18:15,700 in that you would actually exactly expect as you add more cannabinoids for the THCA 173 00:18:15,700 --> 00:18:22,900 or Delta-8 or so on and so forth to diminish because you can actually see legitimate cases of this. 174 00:18:22,900 --> 00:18:28,800 So for example, in Michigan, we had a lot of PSI labs data. 175 00:18:28,800 --> 00:18:36,700 And you could see back in 2015, 2016, they were just fundamentally testing cannabis differently. 176 00:18:36,700 --> 00:18:41,400 I mean, for example, I believe they were even testing it with GC. 177 00:18:41,400 --> 00:18:45,300 So they started off testing just by GC, right? 178 00:18:45,300 --> 00:18:51,800 Because they were mostly just trying to get total THC, total CBD. 179 00:18:51,800 --> 00:18:57,600 And today, right, you move to HTLC, the numbers are going to be fundamentally different, right? 180 00:18:57,600 --> 00:19:02,600 You would still expect the total THC and CBD to be in the same ballpark. 181 00:19:02,600 --> 00:19:04,100 They're still being measured differently. 182 00:19:04,100 --> 00:19:05,700 And then, of course, the time effects. 183 00:19:05,700 --> 00:19:09,200 So yes, over this, it's real interesting data. 184 00:19:09,200 --> 00:19:14,400 I think I shared it with you last week, but I'll make sure that you have a hold of it because the PSI labs data, 185 00:19:14,400 --> 00:19:16,200 I think, is pretty good. 186 00:19:16,200 --> 00:19:19,400 Do we know for certain that nothing was falsified? 187 00:19:19,400 --> 00:19:23,000 No, we kind of have to go off of heuristics. 188 00:19:23,000 --> 00:19:29,400 And from seeing the scientific director speak and whatnot, 189 00:19:29,400 --> 00:19:35,000 I don't have any reasons to believe they're putting out wrong data. 190 00:19:35,000 --> 00:19:40,400 Their data is interesting to observe over time because you see methods change 191 00:19:40,400 --> 00:19:47,400 and you can observe cultivators in Michigan that appears to be getting better at their craft over time. 192 00:19:47,400 --> 00:19:53,600 Oh, yes. And then you raised the interesting question is, and we could actually look at this. 193 00:19:53,600 --> 00:19:56,000 We could do a regression. 194 00:19:56,000 --> 00:20:03,100 One of our explanatory variables would be number of cannabinoids. 195 00:20:03,100 --> 00:20:06,700 So how many cannabinoids are in your panel? 196 00:20:06,700 --> 00:20:10,200 Is it eight? Is it 12? Is it 16? 197 00:20:10,200 --> 00:20:15,900 And then your variable of interest would say be THCA. 198 00:20:15,900 --> 00:20:19,600 And then you could, I guess, try to try to disentangle that and say, 199 00:20:19,600 --> 00:20:29,200 okay, you know, as we add analytes to our panel, do we detect more or less of these various confidence? 200 00:20:29,200 --> 00:20:32,100 You're smiling. Wait, does this sound silly or? 201 00:20:32,100 --> 00:20:35,900 No, no, no, I think it's perfect so long as the data is clean. 202 00:20:35,900 --> 00:20:41,300 As in if you can take out if you're specifically talking about PSI's data 203 00:20:41,300 --> 00:20:44,600 and we believe that their data is accurate, then yes. 204 00:20:44,600 --> 00:20:48,300 But if we were to look at, let's say, all of Washington data, 205 00:20:48,300 --> 00:20:56,000 then we would first have to remove any data from labs that may not be trusted. 206 00:20:56,000 --> 00:20:58,800 Because that would mess up the whole experiment. 207 00:20:58,800 --> 00:21:06,600 And we may not know necessarily how many cannabinoids each of the labs are offering at each time. 208 00:21:06,600 --> 00:21:11,500 I think that analysis is best lab by lab. 209 00:21:11,500 --> 00:21:17,100 Or you could always basically have a lab specific fixed effect. 210 00:21:17,100 --> 00:21:20,100 I don't know if I've explicitly recommended these before, 211 00:21:20,100 --> 00:21:27,500 but you definitely want to have lab specific fixed effects if you're doing analysis for multiple labs. 212 00:21:27,500 --> 00:21:32,500 So, for example, Rick, when you're doing genetic analysis, 213 00:21:32,500 --> 00:21:39,500 say you want to do some prediction on the effects of your strains or this or that, 214 00:21:39,500 --> 00:21:42,100 if you're getting data from a bunch of different, 215 00:21:42,100 --> 00:21:46,400 you may want to get data from multiple labs. 216 00:21:46,400 --> 00:21:52,900 And then if you do so, you may want to basically control for lab by lab variants. 217 00:21:52,900 --> 00:21:58,600 Back to the topic at hand, I think is, what are all the different ways I guess fraud can be committed? 218 00:21:58,600 --> 00:22:01,600 Or it may not necessarily be fraud. 219 00:22:01,600 --> 00:22:07,700 For example, a consideration that I had, I don't know if this is the most likely, 220 00:22:07,700 --> 00:22:11,500 but it could be a co-founding variant. 221 00:22:11,500 --> 00:22:14,100 And I don't want to throw them under the bus because these are the hardest, 222 00:22:14,100 --> 00:22:17,200 in my opinion, the hardest working people at the laboratory. 223 00:22:17,200 --> 00:22:23,900 Or unfortunately, the analysts may have sort of like a pressure like, 224 00:22:23,900 --> 00:22:29,100 oh, the director wants us to have really high lab results. 225 00:22:29,100 --> 00:22:35,100 And they may just be kind of heavy handed, so to speak, not intentionally, 226 00:22:35,100 --> 00:22:40,300 but just kind of introduce bias of their own because they may just say like, 227 00:22:40,300 --> 00:22:46,600 oh, you know, all my peers, all their measurements are coming out at 30%. 228 00:22:46,600 --> 00:22:52,700 Maybe I'll just be a little basically the way the analysts, 229 00:22:52,700 --> 00:22:57,000 the way that would end up happening is you just be maybe a little heavy handed 230 00:22:57,000 --> 00:23:01,100 on the amount that you sample out and you'd be a little light handed 231 00:23:01,100 --> 00:23:04,200 on basically the amount of methanol you would add in. 232 00:23:04,200 --> 00:23:10,700 I don't think that would be the primary way that fraud would be occurring, 233 00:23:10,700 --> 00:23:15,100 but it would just be kind of something to take into consideration is, 234 00:23:15,100 --> 00:23:18,600 unfortunately, like I said, these are hard working people, 235 00:23:18,600 --> 00:23:21,000 may not be getting paid the most. 236 00:23:21,000 --> 00:23:27,200 And so they may just have sort of this pressure to weigh things heavy, so to speak. 237 00:23:27,200 --> 00:23:31,900 But I don't think that's the, you just think it's just something to consider. 238 00:23:31,900 --> 00:23:36,900 And then, of course, just outright fraud, just outright changing the numbers, 239 00:23:36,900 --> 00:23:42,900 Benford's law, which surprisingly, I never really played around with this before. 240 00:23:42,900 --> 00:23:47,100 So Isaac, if you have any insights or maybe one meetup, 241 00:23:47,100 --> 00:23:52,600 if you want to share any luck or work that you've done with this, 242 00:23:52,600 --> 00:23:54,800 I'd be excited to learn more about this. 243 00:23:54,800 --> 00:23:59,400 The final variant here is just, and in fact, 244 00:23:59,400 --> 00:24:07,000 this is I think what people complain the most or I've heard people complain the most about in Oregon, 245 00:24:07,000 --> 00:24:11,300 is based in, in fact, California recently too, 246 00:24:11,300 --> 00:24:17,600 is basically the cultivator would send in a non-random sample 247 00:24:17,600 --> 00:24:20,900 or potentially even an adulterated sample. 248 00:24:20,900 --> 00:24:26,700 The idea is, say in Washington state, you've got to take a random sample from a five pound block. 249 00:24:26,700 --> 00:24:31,900 And I think they'll, cultivators will even get audited. 250 00:24:31,900 --> 00:24:36,000 So the auditor will come by and say, okay, how do you sample from your SOP? 251 00:24:36,000 --> 00:24:41,100 And they say, oh, you know, we'll just take, you know, five random nugs from this five pound lot. 252 00:24:41,100 --> 00:24:46,400 But it's sort of like, okay, well, you know, how do they sample when the auditor is not there? 253 00:24:46,400 --> 00:24:51,200 You know, are they just picking the choicest buds? 254 00:24:51,200 --> 00:25:00,000 In Oregon, there's been claims that people will like sprinkle teeth or even, I've heard people will even, 255 00:25:00,000 --> 00:25:05,600 and this is where I say, so there's a couple of things people will say, oh, well, they'll paint them. 256 00:25:05,600 --> 00:25:11,500 So they'll like get a cannabis concentrate and, you know, lather that on. 257 00:25:11,500 --> 00:25:13,800 Well, then you run into the situation. 258 00:25:13,800 --> 00:25:19,300 Well, we'll say you've got a cannabis flower with concentrate lathered on it. 259 00:25:19,300 --> 00:25:29,300 Well, a good analyst at the laboratory, you know, they're going to look at this under a microscope to look at it for contaminants or, 260 00:25:29,300 --> 00:25:33,300 you know, foreign matter. They'll see, oh, this has been adulterated. 261 00:25:33,300 --> 00:25:36,100 Well, now this analyst is in, right. 262 00:25:36,100 --> 00:25:42,500 Once again, you've got a person who's getting paid, you know, the bare minimum. 263 00:25:42,500 --> 00:25:45,300 Okay, they see the sample. It's been adulterated. 264 00:25:45,300 --> 00:25:54,400 Well, that's going to be a real part, part, part in the pine because it's really no joking matter, 265 00:25:54,400 --> 00:25:59,400 but they're going to be in a sticky situation because, okay, what do you do? 266 00:25:59,400 --> 00:26:09,800 Do you now say this is actually my recommendation is the lab should then fail that sample for performing matter. 267 00:26:09,800 --> 00:26:15,500 They should say, oh, this sample's been adulterated. We're going to fail this. 268 00:26:15,500 --> 00:26:20,000 Well, now that's now in the traceability system. 269 00:26:20,000 --> 00:26:23,500 The state's going to say you what you did. What, what, what? 270 00:26:23,500 --> 00:26:26,500 Yeah, you know, you adulterated this sample. 271 00:26:26,500 --> 00:26:31,000 The cultivator could potentially be in big trouble. 272 00:26:31,000 --> 00:26:34,800 You know, they may lose their license. 273 00:26:34,800 --> 00:26:37,600 They could easily lose their license over something like that. 274 00:26:37,600 --> 00:26:41,800 What seemed like standard practice from the cultivator, right? 275 00:26:41,800 --> 00:26:46,300 They say, oh, all the other cultivators are adulterating our product. 276 00:26:46,300 --> 00:26:49,200 We'll adulterate it. They may lose their license. 277 00:26:49,200 --> 00:26:56,600 And then the lab, once again, is there in a difficult position because I've heard labs say that, 278 00:26:56,600 --> 00:26:59,600 oh, you know, we're not the potency police. 279 00:26:59,600 --> 00:27:06,400 And I kind of cringed at hearing that because, well, they actually kind of are, right? 280 00:27:06,400 --> 00:27:12,100 Because I was thinking about it the other day and it's there's a reason why laboratories have to get accredited. 281 00:27:12,100 --> 00:27:14,000 And, Yosh, you can probably attest to this. 282 00:27:14,000 --> 00:27:20,300 And, Isaac, this is actually the point you raised at the beginning is it's kind of a bit of a privilege, 283 00:27:20,300 --> 00:27:23,100 so to speak, to be a licensed laboratory. 284 00:27:23,100 --> 00:27:29,000 You know, you're there for the public to say, yeah, so there's a lot of responsibility there. 285 00:27:29,000 --> 00:27:33,500 You know, that's, I think, the reason why it's not cheap to start a laboratory 286 00:27:33,500 --> 00:27:36,300 and the reason why you have to go through these accreditations. 287 00:27:36,300 --> 00:27:41,800 Is because you're the ones who do have to make the tough call at the end of the day and say, 288 00:27:41,800 --> 00:27:44,900 yep, these guys said and send an adulterated sample. 289 00:27:44,900 --> 00:27:49,300 It's tough, but we're going to have to now fail you for foreign matter. 290 00:27:49,300 --> 00:27:51,900 They're probably going to take your license away. 291 00:27:51,900 --> 00:27:55,000 You know, that's just a really, really tough call. 292 00:27:55,000 --> 00:27:58,300 And like I said, they may not lose their license, right? 293 00:27:58,300 --> 00:28:02,400 So they depending on the state, maybe they'll get a warning. 294 00:28:02,400 --> 00:28:05,200 Maybe they'll say, OK, we're going to give you a warning. 295 00:28:05,200 --> 00:28:07,000 It may vary by state. 296 00:28:07,000 --> 00:28:08,800 You know, it's a tough call. 297 00:28:08,800 --> 00:28:11,000 Who's responsible? 298 00:28:11,000 --> 00:28:17,100 But I don't think that's something that the laboratory should just pretend they didn't see. 299 00:28:17,100 --> 00:28:19,600 So that's sort of my take on it. 300 00:28:19,600 --> 00:28:25,800 I suggested this in the past where, OK, we found this big ugly problem. 301 00:28:25,800 --> 00:28:30,400 And instead of just stopping there, let's actually offer a solution. 302 00:28:30,400 --> 00:28:34,200 It's not going to be the best solution in the world, but it can be a starting point. 303 00:28:34,200 --> 00:28:42,400 And from what I've seen, a simple, effective, cost-effective way to go about solving this is just say, 304 00:28:42,400 --> 00:28:50,300 you have to keep the samples for 90 days or you have to keep the sample for 60 days. 305 00:28:50,300 --> 00:28:58,500 This imposes a cost on the labs because they'll have to invest in refrigeration or what have you. 306 00:28:58,500 --> 00:29:03,800 But that's a relatively minor cost than some of the alternatives. 307 00:29:03,800 --> 00:29:11,000 And then the idea is, OK, if there's any question later on down the line, 308 00:29:11,000 --> 00:29:13,600 then you can come back and check the sample. 309 00:29:13,600 --> 00:29:20,700 And basically, if something tests in the store at 40% for flour, 310 00:29:20,700 --> 00:29:28,300 you could just go back to the laboratory and just say, OK, you know, can I in fact see this? 311 00:29:28,300 --> 00:29:31,600 Is it obvious that it was adulterated? 312 00:29:31,600 --> 00:29:40,800 If it's not obvious, then, oh, well, you know, and then or then ask the laboratory to do a confirmation run. 313 00:29:40,800 --> 00:29:44,000 But question before I just keep droning on. 314 00:29:44,000 --> 00:29:48,000 I just want to weigh in. Do you mind if I weigh in on last week's next? 315 00:29:48,000 --> 00:29:55,800 So let's say in Massachusetts, we view pre-rolled joints of flour that is tested for that as flour. 316 00:29:55,800 --> 00:30:00,200 It's not separated by testing category. It is that. 317 00:30:00,200 --> 00:30:06,000 And sometimes the stuff that makes pre-rolled joints is a combination of flour and some concentrate. 318 00:30:06,000 --> 00:30:11,800 And it's submitted to us as flour. And then it's definitely altered. 319 00:30:11,800 --> 00:30:16,100 And to me, from, I guess, the ethical standpoint is. 320 00:30:16,100 --> 00:30:19,000 This is what they submitted. This is what we're testing. 321 00:30:19,000 --> 00:30:22,000 This is the results that we're going to give them. 322 00:30:22,000 --> 00:30:28,300 And it accurately says what's in the product that was given to us 323 00:30:28,300 --> 00:30:32,500 and accurately represents the product that consumers will end up buying. 324 00:30:32,500 --> 00:30:39,200 If that product that consumers end up buying is also. 325 00:30:39,200 --> 00:30:44,200 Given the extra whatever was put into the testing sample. 326 00:30:44,200 --> 00:30:54,000 And that if let's say there's a grower that wants to hack the system, wants to send in a sample that will have more. 327 00:30:54,000 --> 00:30:57,900 That will have concentrate tossed in it just for higher results. 328 00:30:57,900 --> 00:31:02,800 If they do it just once, shame on them. They absolutely shouldn't. 329 00:31:02,800 --> 00:31:07,800 And if the results are ridiculous, then I think a secret shopper program would yield that. 330 00:31:07,800 --> 00:31:13,400 OK, the regulators bought a product, they got it tested and the results differ very much. 331 00:31:13,400 --> 00:31:19,600 Between what consumers buy and what the original results were. 332 00:31:19,600 --> 00:31:22,600 And that should lead to an analysis of the data. 333 00:31:22,600 --> 00:31:29,900 Does it show that this is an outlier? Or is this a consistent thing? 334 00:31:29,900 --> 00:31:33,000 Well, I can surely hear you tell how it is. 335 00:31:33,000 --> 00:31:39,400 Exactly. So I think states just need to be conscious and explicitly say how this needs to be handled. 336 00:31:39,400 --> 00:31:47,000 So, for example, in Washington state, they just say if you add anything to it, it's technically a mixed product. 337 00:31:47,000 --> 00:31:50,400 I don't see anything wrong if you want to call it a mixed product. 338 00:31:50,400 --> 00:31:57,200 But I think the idea behind the testing was whatever you send in should be representative of what you sell. 339 00:31:57,200 --> 00:32:03,700 If you send in flowers sprinkled with teeth on it, as long as that's what the label says 340 00:32:03,700 --> 00:32:08,800 and depending on what state you're in, you may call it a mixed product, then by all means. 341 00:32:08,800 --> 00:32:12,000 I mean, in fact, people have really creative products, right? 342 00:32:12,000 --> 00:32:14,900 There's a famous product, the moon rocks. 343 00:32:14,900 --> 00:32:17,700 And that actually is just exactly that. 344 00:32:17,700 --> 00:32:23,200 Just a cannabis flower coated in oil and rolled around in teeth. 345 00:32:23,200 --> 00:32:25,700 So people will specifically buy those. 346 00:32:25,700 --> 00:32:32,600 But that's fine. But if you're selling moon rocks, send in representative moon rocks to be tested. 347 00:32:32,600 --> 00:32:39,400 I don't think it's 100% okay to just say you're selling flour 348 00:32:39,400 --> 00:32:47,200 because I think the big complaint is people just get low quality flour, sprinkle teeth on it. 349 00:32:47,200 --> 00:32:49,500 What they send in isn't what they're selling. 350 00:32:49,500 --> 00:32:51,300 And then Isaac, question? 351 00:32:51,300 --> 00:32:54,900 Yeah, I just want to kind of put in my two cents on this. 352 00:32:54,900 --> 00:32:58,400 It seems to me that there really is two separate issues. 353 00:32:58,400 --> 00:33:05,100 The first is whether the sample that's tested is representative of whatever is being sold. 354 00:33:05,100 --> 00:33:13,100 And the second being the type of product like a pre-roll, as Yasha was saying. 355 00:33:13,100 --> 00:33:21,200 I mean, cannabis market in general is very new and there's always new products being made. 356 00:33:21,200 --> 00:33:27,600 And we have categorization like flour extracts, concentrates, 357 00:33:27,600 --> 00:33:33,000 but pre-roll is blurring the line between flour and extracts, right? 358 00:33:33,000 --> 00:33:39,000 And that's more of a regulatory problem or categorization than representative. 359 00:33:39,000 --> 00:33:45,900 So I just want to separate those two issues as they are rather than having them mixed together. 360 00:33:45,900 --> 00:33:47,700 100% agreed. 361 00:33:47,700 --> 00:33:53,900 And this is once again, I think we said this last week, but it helps to reiterate this week. 362 00:33:53,900 --> 00:34:01,200 That's why we're talking about these things is it just helps to say them out loud and formalize them. 363 00:34:01,200 --> 00:34:04,600 As I shared this piece of policy with you out of Washington State, 364 00:34:04,600 --> 00:34:09,300 a lot of the language either doesn't take product types into consideration 365 00:34:09,300 --> 00:34:14,600 or when they do, they do it in maybe not the most logical fashion, right? 366 00:34:14,600 --> 00:34:21,600 So just the fact that you pointed out that, oh, there's actually two distinct problems going on here is helpful 367 00:34:21,600 --> 00:34:25,600 because now, right, regulators can take that into consideration that, oh, yes, 368 00:34:25,600 --> 00:34:32,400 we need to be explicit about how are different product types being sampled and tested. 369 00:34:32,400 --> 00:34:36,300 You can say, oh, take a representative batch from five pounds of flour. 370 00:34:36,300 --> 00:34:43,500 Well, then also have sampling rules for joints, have sampling rules for mixed products, concentrate, so on and so forth. 371 00:34:43,500 --> 00:34:49,700 And then also, you may want to the states may want to be explicit about sampling. 372 00:34:49,700 --> 00:34:53,300 You know, what does adulteration, what does that mean? 373 00:34:53,300 --> 00:34:56,300 What are the consequences? 374 00:34:56,300 --> 00:34:57,900 Who is responsible? 375 00:34:57,900 --> 00:35:01,700 Because as I was saying, the labs are in a difficult situation, right? 376 00:35:01,700 --> 00:35:02,900 If all the other, right. 377 00:35:02,900 --> 00:35:04,600 So, for example, Massachusetts, right. 378 00:35:04,600 --> 00:35:07,000 And in fact, I just saw this today. 379 00:35:07,000 --> 00:35:11,600 So, right, if all the other labs in Massachusetts, you all want to be on the same page 380 00:35:11,600 --> 00:35:19,900 because just today I saw that ARM Labs in Colorado has decided to close. 381 00:35:19,900 --> 00:35:24,500 And the reason they stated was they simply couldn't stay competitive 382 00:35:24,500 --> 00:35:30,000 because they think other labs in the state aren't following the rules 383 00:35:30,000 --> 00:35:35,400 and they were doing their best to do testing by the book and they just couldn't stay competitive. 384 00:35:35,400 --> 00:35:39,500 And so, Isaac, you mentioned at the beginning, like, you know, you're trying to solve this problem. 385 00:35:39,500 --> 00:35:42,200 And, you know, I applaud you for doing that. 386 00:35:42,200 --> 00:35:43,700 And it matters. 387 00:35:43,700 --> 00:35:50,300 I do think, you know, MCR labs can be profitable in the long term just doing nice ethical testing. 388 00:35:50,300 --> 00:35:55,400 It's just how do you get through the all this short term noise with, you know, 389 00:35:55,400 --> 00:35:59,500 maybe someone sets up shop in there, they're cutting corners. 390 00:35:59,500 --> 00:36:02,400 So how do you survive the short term? 391 00:36:02,400 --> 00:36:05,900 So I love that you're thinking about this and working on it. 392 00:36:05,900 --> 00:36:09,900 Any more thoughts while I think if I have any more final thoughts real quick? 393 00:36:09,900 --> 00:36:16,700 Isaac, did you happen to bring up the question that was asked on Friday with what model would best, 394 00:36:16,700 --> 00:36:22,800 what can be used for identifying anomalies within these kind of data sets? 395 00:36:22,800 --> 00:36:25,800 No, I haven't asked it. 396 00:36:25,800 --> 00:36:32,400 But yeah, I mean, I very much appreciate if the group have any ideas. 397 00:36:32,400 --> 00:36:36,200 Essentially, we're seeing in Massachusetts data, 398 00:36:36,200 --> 00:36:45,300 there is likely a gap of flowers around with around 18 to 19 percent concentration. 399 00:36:45,300 --> 00:36:50,500 And there's a spike of flowers around 20 percent THC concentration. 400 00:36:50,500 --> 00:37:00,300 As I'm sure we all know, flowers with a threshold, flowers above that sells much better. 401 00:37:00,300 --> 00:37:06,700 And I'm just wondering if you're aware of any statistical method 402 00:37:06,700 --> 00:37:13,700 that is a good measurement of this kind of gap? 403 00:37:13,700 --> 00:37:16,200 This is a phenomenal question. 404 00:37:16,200 --> 00:37:20,700 And perhaps when we'll have to chew on more for next week, 405 00:37:20,700 --> 00:37:24,900 I'll just go ahead and give you my hot take right now 406 00:37:24,900 --> 00:37:27,900 and then think about this more thoroughly over the next week 407 00:37:27,900 --> 00:37:30,300 because this seems like this is a pressing issue. 408 00:37:30,300 --> 00:37:35,600 My hot take is, right, I think you were already on a fruitful approach with Bamford's law. 409 00:37:35,600 --> 00:37:41,400 I would like to just reiterate what I was saying at the beginning, 410 00:37:41,400 --> 00:37:46,500 where I think ultimately to for certain predict fraudulent values, 411 00:37:46,500 --> 00:37:50,700 you'll need ones that you know are fraudulent, right? 412 00:37:50,700 --> 00:37:55,200 Because, well, maybe you won't need them for certain, I guess. 413 00:37:55,200 --> 00:37:58,500 But I think that would be helpful is, 414 00:37:58,500 --> 00:38:05,300 okay, what do fraudulent lab results look like compared to non-fraudulent? 415 00:38:05,300 --> 00:38:09,700 And as I was saying, the data set I think that you can get a hold of 416 00:38:09,700 --> 00:38:12,300 are the practice laboratory results 417 00:38:12,300 --> 00:38:17,300 because you know for certain which ones were falsified 418 00:38:17,300 --> 00:38:24,400 and you can compare those to ones that were probably not falsified. 419 00:38:24,400 --> 00:38:27,500 So you can see if there's any systematic difference there. 420 00:38:27,500 --> 00:38:29,600 And that's where Bamford's law could kick in. 421 00:38:29,600 --> 00:38:34,600 So potentially there are digit differences with the falsified ones. 422 00:38:34,600 --> 00:38:39,000 And then finally, this was something that we had hit on in the past 423 00:38:39,000 --> 00:38:42,800 because Rick, once again, getting back to genetics, 424 00:38:42,800 --> 00:38:46,900 we were actually using a some sort of, 425 00:38:46,900 --> 00:38:48,200 I'm not going to get the name right, 426 00:38:48,200 --> 00:38:52,100 but some sort of multivariate difference in means test. 427 00:38:52,100 --> 00:38:55,600 It was the last episode of statistics we did. 428 00:38:55,600 --> 00:38:57,400 But basically we were trying to say, 429 00:38:57,400 --> 00:39:01,000 okay, there's all these plant patents out there 430 00:39:01,000 --> 00:39:03,700 and the way people are patenting their plants 431 00:39:03,700 --> 00:39:07,700 or one of the ways is by chemical composition. 432 00:39:07,700 --> 00:39:09,800 And so we were saying, okay, you know, 433 00:39:09,800 --> 00:39:12,800 could you actually do a difference in means test? 434 00:39:12,800 --> 00:39:18,400 To basically say, okay, this variety, we've tested it 30 times. 435 00:39:18,400 --> 00:39:21,400 We've tested this variety 30 times. 436 00:39:21,400 --> 00:39:24,200 This one is significantly, 437 00:39:24,200 --> 00:39:28,600 this one's chemical composition is significantly different 438 00:39:28,600 --> 00:39:30,800 than this chemical composition. 439 00:39:30,800 --> 00:39:35,300 And that was how we were recommending people go about patenting their plants 440 00:39:35,300 --> 00:39:38,800 where you would just say, okay, we've tested this plant. 441 00:39:38,800 --> 00:39:42,400 It's statistically different than all the existing patents 442 00:39:42,400 --> 00:39:44,900 and it would make a good patent candidate. 443 00:39:44,900 --> 00:39:51,400 You could potentially do a similar sort of difference in means test with samples. 444 00:39:51,400 --> 00:39:53,800 But I think you'd have to get, 445 00:39:53,800 --> 00:39:58,400 think outside of the box about what means you'd be wanting to compare. 446 00:39:58,400 --> 00:40:01,800 And so this is why I floated last week 447 00:40:01,800 --> 00:40:04,700 that you may want to start looking at ratios. 448 00:40:04,700 --> 00:40:08,000 So for example, we looked at it briefly that, okay, 449 00:40:08,000 --> 00:40:13,800 the ratio between total cannabinoids and total terpenes, 450 00:40:13,800 --> 00:40:15,400 I think we looked at this briefly, 451 00:40:15,400 --> 00:40:19,600 but I wanted to say there was a linear relationship between these two, 452 00:40:19,600 --> 00:40:21,800 which was just kind of interesting. 453 00:40:21,800 --> 00:40:25,400 And like I said, we only have briefly explored this. 454 00:40:25,400 --> 00:40:31,100 You know, maybe there's a sample that basically has a weird total cannabinoid 455 00:40:31,100 --> 00:40:33,800 to total THC ratio. 456 00:40:33,800 --> 00:40:39,700 And maybe it also has a weird beta pining to delimiting ratio. 457 00:40:39,700 --> 00:40:43,900 Maybe you'd have, and I think this is kind of what our friend, 458 00:40:43,900 --> 00:40:46,800 John Abrams over at the CESC uses. 459 00:40:46,800 --> 00:40:52,900 He uses a set of ratios to kind of see if lab results look logical. 460 00:40:52,900 --> 00:40:53,700 Ooh, Rick. 461 00:40:53,700 --> 00:40:55,200 Yeah, sorry, I didn't want to cut you off, 462 00:40:55,200 --> 00:40:59,800 but I wanted to agree with what you were saying in terms of like the lab sampling 463 00:40:59,800 --> 00:41:04,500 and having certain profiles or, you know, like the MIMP profile 464 00:41:04,500 --> 00:41:07,800 and attributing that to a specific terpene 465 00:41:07,800 --> 00:41:14,200 and then that to a specific strain or genetic and be able to track that. 466 00:41:14,200 --> 00:41:17,700 Also with things like hoplite and viral, 467 00:41:17,700 --> 00:41:21,800 it's important now to also for clean cuts 468 00:41:21,800 --> 00:41:26,800 or things coming out of tissue culture labs to establish a chain of custody. 469 00:41:26,800 --> 00:41:30,200 So having a way to do that as well is all interesting. 470 00:41:30,200 --> 00:41:34,700 But right now, none of that data is released for like the hoplite and testing. 471 00:41:34,700 --> 00:41:38,600 I noticed a bunch of startup testing labs 472 00:41:38,600 --> 00:41:41,500 where they can send you at home or they'll have people come on site. 473 00:41:41,500 --> 00:41:46,200 There seems to be kind of a lot of data streams for where that's going. 474 00:41:46,200 --> 00:41:51,700 And so right now it's not readily available or in a database that I'm aware of. 475 00:41:51,700 --> 00:41:54,800 But that's the sort of stuff that interests me. 476 00:41:54,800 --> 00:41:58,700 And then also, you know, if you're working with, you know, 477 00:41:58,700 --> 00:42:02,700 trusted data in terms of testing of those genetics as well, 478 00:42:02,700 --> 00:42:07,600 collecting information from the facilities such as light spectrum, 479 00:42:07,600 --> 00:42:12,700 medium that they use, different types of fertilizers or whatever 480 00:42:12,700 --> 00:42:17,500 and how that may express down the line with different cannabinoid profiles 481 00:42:17,500 --> 00:42:22,300 or, you know, understanding that relation as well or terpene profiles. 482 00:42:22,300 --> 00:42:27,600 I love it, Rick. And you were right at the frontier and you brought up many good points. 483 00:42:27,600 --> 00:42:31,800 And I think what you hit on is what's the starting point? 484 00:42:31,800 --> 00:42:37,100 Get the data, organize it, curate it, and then of course, get around to analyzing it. 485 00:42:37,100 --> 00:42:38,900 You're right on the path, right? 486 00:42:38,900 --> 00:42:42,000 And that is right now there is poor data. 487 00:42:42,000 --> 00:42:44,400 So first things first, get it. 488 00:42:44,400 --> 00:42:49,500 And so I love that. And then you raise issues like, oh, hoplite and viroid. 489 00:42:49,500 --> 00:42:55,800 And this is why I was saying, well, there's a million co-founding effects going on, right? 490 00:42:55,800 --> 00:43:01,200 And so, for example, oh, maybe you're looking at total cannabinoids to terpenes, 491 00:43:01,200 --> 00:43:04,500 and that's a really good ratio. And one sample's off. 492 00:43:04,500 --> 00:43:08,300 Well, maybe we're not taking into consideration hoplite and viroid. 493 00:43:08,300 --> 00:43:13,000 Maybe the viroid interferes with these standard ratios. 494 00:43:13,000 --> 00:43:17,900 So maybe you think, oh, maybe somebody's falsifying results for this one sample. 495 00:43:17,900 --> 00:43:23,100 But, oh, you go take a closer look and it's like, oh, maybe these plants had trace amounts of... 496 00:43:23,100 --> 00:43:25,900 Or maybe these plants were slightly infected with the virus 497 00:43:25,900 --> 00:43:32,000 and they just weren't producing terpenes normally or something of that sort. 498 00:43:32,000 --> 00:43:34,000 Once again, it's all investigation. 499 00:43:34,000 --> 00:43:37,600 That's part of what's interesting about being a data scientist, right? 500 00:43:37,600 --> 00:43:39,400 We wear many hats, right? 501 00:43:39,400 --> 00:43:42,600 As a good researcher, we have to kind of investigate these. 502 00:43:42,600 --> 00:43:45,200 You know, don't jump to conclusions. 503 00:43:45,200 --> 00:43:48,500 But I think we've got some fruitful things to work on. 504 00:43:48,500 --> 00:43:51,900 So, Isaac, I love that you've pointed this problem out, 505 00:43:51,900 --> 00:43:54,900 because as I said, you know, we're all about solutions here. 506 00:43:54,900 --> 00:43:57,700 So let's, you know, put our minds together 507 00:43:57,700 --> 00:44:02,000 and see if we can't think about ways to help the laboratories out. 508 00:44:02,000 --> 00:44:03,700 I think we can use statistics to help. 509 00:44:03,700 --> 00:44:07,000 Well, I think we've got plenty to work on. 510 00:44:07,000 --> 00:44:13,700 So for Rick, we're going to be looking at strain names just to try to find crosses, 511 00:44:13,700 --> 00:44:17,900 because lineage is something that I've been curious about looking at. 512 00:44:17,900 --> 00:44:21,700 Just basically seeing, oh, can we trace all the hazes back? 513 00:44:21,700 --> 00:44:26,000 I'm like working on scraping data from a few different locations. 514 00:44:26,000 --> 00:44:35,100 There's a few sites out there that keep pretty detailed tracking of like a strain and its lineage. 515 00:44:35,100 --> 00:44:39,900 And there's also Reddit, where there's individual subreddits with posts 516 00:44:39,900 --> 00:44:44,500 that you can pull strain names and other stuff out of. 517 00:44:44,500 --> 00:44:51,900 So I'll try to contribute to the GitHub and make sure that anything that I collect will be available there also. 518 00:44:51,900 --> 00:44:52,900 Phenomenal, Rick. 519 00:44:52,900 --> 00:44:57,800 Well, try to come back next week, because genetics and lineages on the agenda, 520 00:44:57,800 --> 00:45:02,100 somehow we got pushed back from today, but definitely next week. 521 00:45:02,100 --> 00:45:06,400 And then we'll also end up looking at consumers in the next week, 522 00:45:06,400 --> 00:45:14,100 because last week we just did a back of the envelope estimate of how many consumers there were in Washington state. 523 00:45:14,100 --> 00:45:19,300 But we can do a much better job of that, given some data that's out there. 524 00:45:19,300 --> 00:45:23,100 So I was thinking, and we can even do some consumer analysis. 525 00:45:23,100 --> 00:45:24,600 So those are the things that are coming up. 526 00:45:24,600 --> 00:45:31,700 We'll get to the bottom of this lab shopping for help Isaac out, help good old NCR labs out. 527 00:45:31,700 --> 00:45:35,600 So we'll help out as much as we can. 528 00:45:35,600 --> 00:45:39,300 Then, of course, genetics and lineage and cannabis consumers. 529 00:45:39,300 --> 00:45:41,900 So those are the topics coming up in the coming weeks. 530 00:45:41,900 --> 00:45:43,700 It's going to be a fun time. 531 00:45:43,700 --> 00:45:46,500 So we'll end the year out strong. 532 00:45:46,500 --> 00:45:51,100 Just like to thank you all for helping advance cannabis science. 533 00:45:51,100 --> 00:45:55,600 Couldn't do it without you, your eyes, your ears, your brilliant minds. 534 00:45:55,600 --> 00:46:01,800 So thank you all for coming together to the Cantlydix Cannabis Data Science Meetup. 535 00:46:01,800 --> 00:46:07,600 So thank you.