1 00:00:00,000 --> 00:00:10,240 Welcome to the cannabis data science meetup group. Glad to have you all here in for a 2 00:00:10,240 --> 00:00:17,520 treat today. Gonna get our hands on some data and picking up from last week be looking at 3 00:00:17,520 --> 00:00:24,000 Benford's law. This was new to me. Our good friend Isaac at MCR labs brought this up as 4 00:00:24,000 --> 00:00:30,400 an interesting tool that we could perhaps use a statistical tool. Did a bit of research. It looks 5 00:00:30,400 --> 00:00:38,560 promising. Got some great statisticians behind it and so wanted to share with you what I've learned, 6 00:00:38,560 --> 00:00:44,960 how we can apply it to some data and get your thoughts about moving the ball forward. That's 7 00:00:44,960 --> 00:00:50,640 the plan for today but just to go ahead and you know give five minutes let everybody say oh you 8 00:00:50,640 --> 00:00:57,040 know what you may hope to get out of the group. I'm gonna start in reverse order today that way 9 00:00:57,760 --> 00:01:03,520 our new friend here Christina doesn't have to be put on the spot. Candice excited to see you today. 10 00:01:04,160 --> 00:01:10,080 Love the hat. Would love to hear about your adventures in the past week. If you were able 11 00:01:10,080 --> 00:01:17,040 to get any data science work done and what you're looking forward to in the coming weeks to finish 12 00:01:17,040 --> 00:01:24,800 off the year strong. Well that's a doozy. So yes so I'm going to be working on parsing a new lab 13 00:01:25,440 --> 00:01:35,600 and also too I'm gonna be putting up my presentations, my plots with Seaborn and 14 00:01:35,600 --> 00:01:40,400 I've been checking out two the YouTube videos that you posted Keegan. That's really great. I did get 15 00:01:40,400 --> 00:01:47,920 to Florida and I did go to Kana MD and so I'm renewed. It was pretty interesting I was talking 16 00:01:47,920 --> 00:01:53,600 to my doctor about some of the things that I was learning with MCR Labs and Jeff Haas and over at 17 00:01:53,600 --> 00:02:01,360 Harvard. We had an interesting discussion and but he also gave me some good leads too for some good 18 00:02:01,360 --> 00:02:06,560 flower in Florida and because it's kind of interesting that you know I joined some user 19 00:02:06,560 --> 00:02:15,360 groups and actually somebody broke open a fresh product from the state of Florida and they showed 20 00:02:15,360 --> 00:02:21,120 me pictures. I have copies of the pictures. It has mold in it you know it's like visible like you can 21 00:02:21,120 --> 00:02:28,640 like pull it apart and I don't know it's uh but anyway so I'm going to be getting some COAs from 22 00:02:28,640 --> 00:02:36,080 Florida also. Well it's not fantastic but it's quite interesting to hear you on the ground in 23 00:02:36,080 --> 00:02:43,200 Florida telling it how it is and mold is an interesting thing because from being in the 24 00:02:43,200 --> 00:02:50,960 laboratory and in fact we'll look at this today not all analyses are taken as seriously as others 25 00:02:51,520 --> 00:02:59,840 unfortunately. Coming at it in an economic lens people have to prioritize and a lot of the priority 26 00:02:59,840 --> 00:03:09,040 goes towards measuring cannabinoids really really accurately and tests like micro detection, 27 00:03:09,040 --> 00:03:16,080 microtoxin screening that may get pushed to the back burner. However from seeing some of the 28 00:03:16,080 --> 00:03:23,920 unintended consequences one thing that is a little concerning to me is that some large-scale 29 00:03:23,920 --> 00:03:32,320 cultivators are seeing this as a cost which I pointed out that it was a while back right you 30 00:03:32,320 --> 00:03:38,640 have to factor in the probability that you may fail a test and that's a cost and if you don't 31 00:03:38,640 --> 00:03:45,920 factor it in you could run into the red with your business. However some of the large cultivators are 32 00:03:45,920 --> 00:03:54,640 thinking okay how can we prevent failing say micro testing and there are various treatments out there 33 00:03:54,640 --> 00:04:00,720 as we know with a lot of things you know how well studied are some of these things and so just to be 34 00:04:00,720 --> 00:04:07,840 to be frank some of the ones going from from least concerning to more concerning pressure 35 00:04:07,840 --> 00:04:14,640 oxygen deprivation perhaps to treating the product with ozone and then Candice I think you mentioned 36 00:04:14,640 --> 00:04:20,640 at one point radiation and those are the ones that are slightly more concerning to me while they may 37 00:04:20,640 --> 00:04:29,360 be say chemically sound I don't know I just wouldn't wouldn't mind a bit further research myself 38 00:04:29,360 --> 00:04:37,440 so that can be an unintended consequence of mandated micro testing but as Candice you pointed out 39 00:04:37,440 --> 00:04:46,400 mold is a thing right and no one wants moldy cannabis flower the consumer may want a reasonable 40 00:04:46,400 --> 00:04:53,040 expectation that their products won't have mold in them so what to do about that. I forget the name 41 00:04:53,040 --> 00:04:58,240 of it the disease but it's like some type of disease that I think people get in agriculture 42 00:04:58,240 --> 00:05:03,920 and from breathing in mold I'll have to nail it down you know I mean it's a new industry and it's 43 00:05:03,920 --> 00:05:09,360 awesome cannabis is amazing right compared to big pharma and people can really get off a lot of 44 00:05:09,360 --> 00:05:16,080 drugs Keegan right and with cannabis but then also too you know it's like yeah they really do need I 45 00:05:16,080 --> 00:05:20,400 think they need transparency we need open data sets in Florida and Massachusetts that's what we 46 00:05:20,400 --> 00:05:25,760 need so that the watch dogs right can help keep an eye on the system like Jeff Haasen says right 47 00:05:25,760 --> 00:05:31,680 well the federal government right the FDA do we really we us kids in the states we need to play 48 00:05:31,680 --> 00:05:36,960 good and be transparent and have the data out there so that the federal government doesn't turn 49 00:05:36,960 --> 00:05:42,800 around and turn it into tobacco or you know because it's funny because like my first review with 50 00:05:42,800 --> 00:05:48,400 pure leaf that's still up there a year ago talks about how I'm getting this nice organic cannabis 51 00:05:48,400 --> 00:05:54,160 and stuff right and wow what I've learned right because that's really how it sold to us patients 52 00:05:54,160 --> 00:05:59,280 you know and and I get it because I was kind of doing the same thing you know that review is still 53 00:05:59,280 --> 00:06:06,240 up there I don't know it's it's just so cool just to know right to just be aware of you know possible 54 00:06:06,240 --> 00:06:12,000 health problems and then too I don't know how they would fix that because uh you know what I 55 00:06:12,000 --> 00:06:18,240 was saying in that picture looked just like uh when I put the predator mics on the uh for my 56 00:06:18,240 --> 00:06:23,760 spider mics that somehow came in on my hydrogen peroxide clone right you raised a couple good 57 00:06:23,760 --> 00:06:29,920 points that have sparked a couple thoughts but before I get to those anyway I'm taking the floor 58 00:06:30,640 --> 00:06:36,560 we'll get back to you Candice I just wanted to let Isaac chime in wouldn't mind hearing your thoughts 59 00:06:36,560 --> 00:06:43,360 on this this microbe yeah I just want to add something uh from the lab's perspective I mean 60 00:06:43,360 --> 00:06:50,080 well each state all require an array of testing right heavy metal uh more uh eastern mold and 61 00:06:50,080 --> 00:06:58,320 other bacterias and but I can attest that uh east mode is among one of the most challenging 62 00:06:58,320 --> 00:07:04,800 essays for clients to pass even if they are huge dispensaries I would say at least half of them 63 00:07:05,920 --> 00:07:13,520 do struggle with the issue so it's uh I'm not surprised that you're finding uh uh flowers with 64 00:07:13,520 --> 00:07:20,480 literally visible mold on them I mean that I've heard that happen from time to time and also 65 00:07:20,480 --> 00:07:28,720 from another perspective uh is the health of the worker if the growing room is contaminated by 66 00:07:28,720 --> 00:07:35,600 uh mold or some kind of bacteria uh the workers are at risk of getting respiratory diseases 67 00:07:35,600 --> 00:07:42,480 as a matter of fact uh last year there is a uh worker in one of the dispensaries unfortunately 68 00:07:42,480 --> 00:07:51,280 passed away and the investigation is still ongoing but uh the direction is toward the link between the 69 00:07:51,280 --> 00:07:58,480 mold contamination in their grow room and her unfortunate passing away so it's also to the 70 00:07:58,480 --> 00:08:04,880 uh workers in the industry you're Christina are you going to chime in before I ramble on um I guess 71 00:08:04,880 --> 00:08:10,560 what concerns me this is about concerns right because what concerns me is the um some of the 72 00:08:10,560 --> 00:08:18,080 some of the strains are getting contaminated I don't know if by chemicals or another strand that 73 00:08:18,080 --> 00:08:26,080 kind of it doesn't go well with the strand like if I were to buy an ACDC right and it's not 74 00:08:26,080 --> 00:08:33,760 completely clean the strain for me it's very important because I have epilepsy and terpene 75 00:08:33,760 --> 00:08:43,520 terpenes they matter to me because one terpene can make a big change in my um my body you know 76 00:08:43,520 --> 00:08:49,120 and I don't want to get a seizure because of a strand being contaminated and um um I live in 77 00:08:49,120 --> 00:08:54,960 Seattle like the Seattle area there's this little shop where I bought a strand and it was indeed 78 00:08:54,960 --> 00:09:02,880 contaminated I got I started to see like weird things and it was just not a fun moment so like 79 00:09:02,880 --> 00:09:11,200 I don't know if that's common or not you're spot on in in the right place because this is the exact 80 00:09:11,200 --> 00:09:17,120 topic at hand that we're talking about in that when you go to a retailer what are they going to 81 00:09:17,120 --> 00:09:24,240 be selling to you they're going to have the the strains labeled oh ACDC or this or that um and 82 00:09:24,240 --> 00:09:31,360 then they may have some of the chemical compositions labeled you've learned that okay ACDC works well 83 00:09:31,360 --> 00:09:38,080 for you and what we're trying to do is tie some of these real life situations with the data and the 84 00:09:38,080 --> 00:09:46,800 best lead that we have is why is ACDC effective for you perhaps it has something to do with the 85 00:09:46,800 --> 00:09:55,760 unique chemical composition of ACDC so we want to get a nice measure of those chemicals because 86 00:09:55,760 --> 00:10:02,560 we're so at the frontier essentially no one consuming cannabis except for maybe a small 87 00:10:02,560 --> 00:10:08,880 minority of people actually know the milligrams of various chemical compounds that they're 88 00:10:08,880 --> 00:10:16,480 ingesting luckily you're in Seattle so there are good testing laboratories in Washington State 89 00:10:16,480 --> 00:10:25,200 and so what you may want to do in fact in Washington State it is required that you may want to ask for 90 00:10:25,200 --> 00:10:33,920 the certificate of analysis when you when you purchase a product then you can keep track of the 91 00:10:33,920 --> 00:10:40,720 chemicals that you're ingesting it would be awesome if the product was tested for terpenes 92 00:10:40,720 --> 00:10:47,360 that's not required but some people do test their products for terpenes that that's kind of concerning 93 00:10:47,360 --> 00:10:53,040 concerning aren't the the products supposed to already come with the terpenes is if that's what 94 00:10:53,040 --> 00:11:01,920 i'm actually purchasing well you raise a awesome point that wasn't really taking into consideration 95 00:11:01,920 --> 00:11:08,240 until late in the game so long story short when cannabis was first legalized right so it was 96 00:11:08,240 --> 00:11:14,400 permitted in Washington State back in 2012 and so originally people were thinking what are the main 97 00:11:14,400 --> 00:11:24,000 psychoactive components those are going to be thc and cbd well since 2012 it's been a decade 98 00:11:24,000 --> 00:11:30,400 10 years of scientific research people have discovered more and more compounds and it's 99 00:11:30,400 --> 00:11:36,960 looking like other compounds may have some significance right maybe not all of them right 100 00:11:36,960 --> 00:11:45,680 maybe chlorophyll may not necessarily have a good or a bad effect but people are discovering oh 101 00:11:45,680 --> 00:11:51,840 there's beta-cariofoli and these also may have effects last comments and then i'll let Isaac 102 00:11:51,840 --> 00:11:56,640 jump in and i was just going to say certain states are starting to realize that we may also 103 00:11:56,640 --> 00:12:03,680 want to measure terpenes i think Oklahoma they're mandated it was sort of an unusual mandate but 104 00:12:03,680 --> 00:12:10,240 they are tested there and before i keep talking i would love to get Isaac love to hear what you 105 00:12:10,240 --> 00:12:17,600 have to say yeah thanks i just want to add that uh dreams like gg4 because cannabis has been 106 00:12:17,600 --> 00:12:26,000 underground for such a long time and the readings are all very messy and and and it's almost 107 00:12:26,000 --> 00:12:33,200 impossible to trace that to the original gene so there are a lot of dispensaries or seed banks they 108 00:12:33,200 --> 00:12:42,640 can claim that their flower is gg4 or if it's blue dream but they're really we don't really have a 109 00:12:42,640 --> 00:12:49,840 reference or a standard or certification okay this is the gg4 phenotype i've actually done similar 110 00:12:49,840 --> 00:12:57,040 analysis of uh using uh principal component analysis just to plot of for example all the 111 00:12:57,040 --> 00:13:06,240 gg4s we receive from different growers and they do appear uh triable so uh one client or one 112 00:13:06,240 --> 00:13:13,280 dispensary's gg4 is likely to be different from another client that's very difficult for 113 00:13:14,000 --> 00:13:20,320 consumers like you to know exactly what what you're purchasing and i think this leads naturally to 114 00:13:20,320 --> 00:13:27,760 what keegan was talking about to measure or to get exact chemical profile of a strain and then 115 00:13:27,760 --> 00:13:35,520 using this more scientific categorization rather than relying on a name which at the moment any 116 00:13:36,160 --> 00:13:43,680 seed bank can claim that their seed is this strain but they might not i'll just keep piggybacking on 117 00:13:43,680 --> 00:13:49,440 that in that let's say strain names the best we have it's a far far from perfect tool but like 118 00:13:49,440 --> 00:13:56,240 you said it's at least getting you on the right track so okay acdc the strain appears to have a 119 00:13:56,240 --> 00:14:01,760 certain effect for you and from my understanding of this strain that it could even be a what's 120 00:14:01,760 --> 00:14:12,320 called a type two or where it's almost an equal thc to cbd ratio it's very equal exactly so i think 121 00:14:12,320 --> 00:14:18,000 there could quite easily be something going on with the the unique chemical composition of that 122 00:14:18,000 --> 00:14:22,640 we're still we're working on this we're going to get to the bottom of this because as isaac said 123 00:14:22,640 --> 00:14:28,960 and it's real peculiar because there's a real disconnect between say the laboratories and 124 00:14:28,960 --> 00:14:35,200 the breeders and maybe there's not as much disconnect as one may think um it may it may 125 00:14:35,200 --> 00:14:40,880 be more of a show all the breeders right isaac said so if you're the breeders are selling their 126 00:14:40,880 --> 00:14:47,440 seeds by strain name and this is something that i wanted to get into what maybe this week but it 127 00:14:47,440 --> 00:14:53,760 was looking like next week's going to be the week is actually looking at some of the the seed vendors 128 00:14:53,760 --> 00:14:59,600 and looking at their history because there's a really really cool history there and all of 129 00:14:59,600 --> 00:15:07,200 these strain names have cool histories but the kind of names and as isaac was saying they're not 130 00:15:07,200 --> 00:15:16,400 set in stone right anyone can grow any weed any cannabis plant and just and call it acdc so that's 131 00:15:16,400 --> 00:15:25,280 the trouble behind it the interesting part is perhaps there is a variety that tends to have 132 00:15:25,280 --> 00:15:32,240 equal distributions of thc and cdd perhaps people kind of want to call that acdc at the end of the 133 00:15:32,240 --> 00:15:38,800 day i think what you want to be concerned about is what you're ingesting and and those will be 134 00:15:39,760 --> 00:15:45,840 the chemical compounds you may actually end up finding another strain and so for example 135 00:15:45,840 --> 00:15:52,560 you know there's canatonic or this or that yeah i like canatonic exactly that also have a similar 136 00:15:52,560 --> 00:15:58,880 chemical composition that was with some of the work we were doing earlier this summer with product 137 00:15:58,880 --> 00:16:05,520 recommendations we're saying oh you know if you gravitate towards this one strain then you may 138 00:16:05,520 --> 00:16:13,920 just want to find a strain that's as chemically similar as possible okay it's imperfect but at 139 00:16:13,920 --> 00:16:20,640 the moment it's you know the best we can do until we can get better hold of this data you know get 140 00:16:20,640 --> 00:16:27,200 it into your hands let you start learning about what works well for you it'll be a slow process 141 00:16:27,200 --> 00:16:33,600 but hopefully there'll be a feedback loop where you know maybe certain growers will start growing 142 00:16:34,160 --> 00:16:40,800 specific cultivars with specific chemical compositions and then the retailers know how 143 00:16:40,800 --> 00:16:48,080 to label this to get it to you to get the intended effect i have one more question well i have a bunch 144 00:16:48,080 --> 00:16:54,720 of questions but on that i'm looking for new strands i know there's one called corazón and 145 00:16:54,720 --> 00:17:04,480 that's from yerbabuena farms up in oregon and one of the parent strands strains is acdc and the other 146 00:17:04,480 --> 00:17:13,200 strain is unknown there's not a lot like research done to it yet but is there a way that you guys 147 00:17:13,200 --> 00:17:21,680 have connections as to how to look for these strains that are not well established yet like i 148 00:17:21,680 --> 00:17:27,760 want to get a hold of this strain but i don't know how right so i just sent in the chat the best kind 149 00:17:27,760 --> 00:17:36,000 of seed genetic website that i have uh give that a try like go directly to the website and request 150 00:17:36,000 --> 00:17:42,560 it there no the website is like a database so you can search for different strains and see their 151 00:17:42,560 --> 00:17:49,520 histories if that's part of the archive what's the website called i would love it thank you 152 00:17:49,520 --> 00:17:59,840 uh en.seedfinder.eu perfect thank you i really appreciate it it's not the most polished looking 153 00:17:59,840 --> 00:18:08,400 website but it has a wealth of information that overwhelms everyone that is okay all right i'll 154 00:18:08,400 --> 00:18:15,200 look into that thank you oh make sure to to come back to this christina since this this is interesting 155 00:18:15,200 --> 00:18:20,880 i'll try to just run through this super super fast for isaac what's it while you're here just wanted 156 00:18:20,880 --> 00:18:28,720 to to share this book came in the mail hemp diseases and pests so this is a an old one it's 157 00:18:28,720 --> 00:18:34,720 interesting right so right the consumers are often concerned about effects but the cultivators are 158 00:18:34,720 --> 00:18:41,200 often concerned about diseases and canons are plagued by many diseases so you know these molds 159 00:18:41,200 --> 00:18:46,320 are a real concern they're of a top concern to cultivators i think we'll want to to talk more 160 00:18:46,320 --> 00:18:53,120 about that i'll i'll run through this super super fast last week isaac brought up benford's law 161 00:18:53,120 --> 00:19:01,040 which is basically when you look at data sets coincidentally in it it's not necessarily 162 00:19:01,040 --> 00:19:08,560 coincidentally i think it's a statistical phenomena different digits occur at certain probabilities 163 00:19:08,560 --> 00:19:14,480 and they follow a certain distribution and so here's the distribution you know one it will appear 164 00:19:14,480 --> 00:19:19,600 around 30 percent of the time we don't have to get too much into the math but you're welcome to 165 00:19:19,600 --> 00:19:28,480 so this was noted way back in the 1800s and then really made concrete in the early 1900s by frank 166 00:19:28,480 --> 00:19:36,320 benford so i'll let you read about this his main contribution was getting many different data sets 167 00:19:36,320 --> 00:19:42,000 and analyzing all these different data sets that made me think of the cannabis data science group 168 00:19:42,000 --> 00:19:48,720 how we're getting many different data sets and applying statistics to them okay what's going to 169 00:19:48,720 --> 00:19:59,120 be required for this we actually want a wide range of numbers if you have a narrow range it's going 170 00:19:59,120 --> 00:20:06,960 to be difficult to show this so so we'll just kind of point that out now and then what are some cool 171 00:20:06,960 --> 00:20:14,960 applications of this if you don't know him halvarian is a famous economist statistician 172 00:20:14,960 --> 00:20:21,280 and he was the one who put forward that this could potentially be used to detect fraud isaac pointed 173 00:20:21,280 --> 00:20:28,160 out potential well not potentially there are cases of laboratory fraud in the cannabis industry and 174 00:20:28,160 --> 00:20:37,280 we would want some mechanism for detecting if a lab result is fraudulent or perhaps if it's probable 175 00:20:37,280 --> 00:20:44,800 that it's fraudulent just from reading the rest of this my main takeaway is that this is a good 176 00:20:44,800 --> 00:20:52,960 mechanism to know what looks suspect and where you may want to look further but i don't know if it's 177 00:20:52,960 --> 00:20:59,280 definitive proof but i think it can be a useful tool about where you may want to look further 178 00:20:59,280 --> 00:21:05,280 the last thing i wanted to point out we actually are running into the problem actually here i'll 179 00:21:05,280 --> 00:21:10,400 point this i'll i'll talk more about this here here momentarily so that's benford's law in a 180 00:21:10,400 --> 00:21:18,320 nutshell different digits appear at different frequencies and if digits don't appear at that 181 00:21:18,320 --> 00:21:25,120 frequency then one may be suspect as to the underlying distribution that's generating the 182 00:21:25,120 --> 00:21:33,360 data so there may not be a nice random underlying generation but this is my really really crude 183 00:21:33,360 --> 00:21:37,280 first take so isaac please feel free to correct me at any point but i want to go ahead and get 184 00:21:37,280 --> 00:21:42,800 through this data to go ahead and kind of get you out of here on time so long story short we have 185 00:21:42,800 --> 00:21:55,680 lab results from washington state in 2022 1.6 million compounds were measured so for example 186 00:21:55,680 --> 00:22:04,480 what does a compound look like this was somebody who tested for microbes this sample actually 187 00:22:04,480 --> 00:22:15,440 failed for microbes it looks like we were interested in say total thc here is the distribution of 188 00:22:15,440 --> 00:22:25,520 thc in cannabis flower in 2022 and this is the phenomena that isaac was pointing out in that 189 00:22:25,520 --> 00:22:35,200 presumably this mode is flower samples and it doesn't appear perfectly normally distributed 190 00:22:35,200 --> 00:22:43,360 and it perhaps may even be skewed towards above 20 i think there's better examples of this but 191 00:22:44,000 --> 00:22:49,040 but that's sort of what we claim just to look at a couple more analytes there's a distribution of 192 00:22:49,040 --> 00:22:58,960 cbd here is your distribution of moisture content and this is more applicable in other states where 193 00:22:58,960 --> 00:23:04,960 lab results are corrected for moisture content i'm not certain if they are in washington state or not 194 00:23:04,960 --> 00:23:12,080 you're really looking for anomalies or non-normal distributions at this stage and one thing i'd like 195 00:23:12,080 --> 00:23:19,520 to point out is you've got this curious mode at 10 and so it looks like for whatever reason some 196 00:23:19,520 --> 00:23:25,200 laboratory may just be reporting just 10 for their moisture content whereas other labs may 197 00:23:25,200 --> 00:23:33,040 be a bit more granular and then water activity this is sort of a measure of shelf stability 198 00:23:33,040 --> 00:23:44,480 and curious thing here is right the limit is 0.65 and this may just be just how the distribution 199 00:23:44,480 --> 00:23:50,640 shook out but you know one may kind of wonder if you know things are kind of being you know slid 200 00:23:50,640 --> 00:23:58,960 under or the 0.65 limit but once again as far as the distribution goes it's looks fairly normal 201 00:23:58,960 --> 00:24:06,800 now quick application of benford's law benford's law requires us to span at least a 10-digit range 202 00:24:06,800 --> 00:24:14,640 so we could do it with thc but it would be tough we wouldn't meet the assumptions required for 203 00:24:14,640 --> 00:24:23,040 benford's law with the other analytes i don't know if this is acceptable but what i did was i just 204 00:24:23,040 --> 00:24:32,560 grabbed the first decimal place for example if you look at thc you'll see most people are reporting 205 00:24:32,560 --> 00:24:40,640 whole numbers so zero is going to occur a lot and then i was also just looking at the first decimal 206 00:24:40,640 --> 00:24:46,720 place here once again this is sort of my quick and dirty attempt at this i don't know if this is 207 00:24:46,720 --> 00:24:53,440 an unacceptable way to implement benford's law but if you look at the digit occurrences 208 00:24:53,440 --> 00:25:02,080 one through nine you do curiously see the decreasing probability of digits it doesn't look 209 00:25:02,080 --> 00:25:08,160 like you know they decrease in probability at quite the same rate as benford observed 210 00:25:08,160 --> 00:25:13,600 but it is interesting that you you see this distribution because if you look at some of the 211 00:25:13,600 --> 00:25:21,840 other analytes well check out cbd cbd diminishes at even further rate i think this is a first 212 00:25:21,840 --> 00:25:30,640 interesting observation is that thc and cbd for whatever reason have different distributions of 213 00:25:30,640 --> 00:25:38,400 digits but they are both are are diminishing which is interesting if you look at moisture content 214 00:25:38,400 --> 00:25:47,920 you'll see almost a uniform distribution so if every digit occurs at the same probability 215 00:25:47,920 --> 00:25:54,320 that would be called a uniform distribution and it would essentially imply that the results are 216 00:25:54,320 --> 00:26:01,200 random i don't know what's going on with moisture content but this looks like it requires further 217 00:26:01,200 --> 00:26:08,800 investigation because it almost just looks like people are just randomly selecting moisture 218 00:26:08,800 --> 00:26:15,680 content or at least the data generation process maybe we may need to think further about the data 219 00:26:15,680 --> 00:26:23,200 generation process and then even more curiously with water activity once again this could just be 220 00:26:23,920 --> 00:26:30,320 something about the data generation process it could just be something unique to cannabis flower 221 00:26:30,320 --> 00:26:38,240 but once again you don't observe benford's law with the digits you see the digits increasing 222 00:26:38,240 --> 00:26:43,920 in probability to four and then decreasing in probability once again this may be a violation 223 00:26:43,920 --> 00:26:52,560 of benford's law so whereas thc and cbd which curiously we thought oh maybe those would be the 224 00:26:52,560 --> 00:26:59,280 ones that were more likely to be falsified curiously and i kind of brought this up at the beginning 225 00:26:59,280 --> 00:27:06,400 that people are putting such great focus into measuring thc and cbd then maybe they actually 226 00:27:06,400 --> 00:27:15,440 are measuring thc and cbd well and maybe little emphasis is put on moisture and water activity 227 00:27:15,440 --> 00:27:21,600 and they may just be just writing down random numbers and when a human writes down a random 228 00:27:21,600 --> 00:27:29,600 number then the digits won't follow benford's law and so that was just curious but that was the 229 00:27:29,600 --> 00:27:37,280 analysis of benford's law this is a start and i would just like to point out that you may need to 230 00:27:37,280 --> 00:27:45,120 take the log of these numbers so so that's a downside and as i pointed out i'm using the first 231 00:27:45,120 --> 00:27:54,400 decimal place and that's also perhaps imperfect but i thought this was curious so i'll kind of 232 00:27:54,400 --> 00:28:02,400 let people think about that what if this is actually one lab that's just doing this but if 233 00:28:02,400 --> 00:28:09,920 you look at all the different laboratories they all sort of have this weird pattern for water 234 00:28:09,920 --> 00:28:20,720 activity long story short is i don't think it's a specific laboratory that is doing this it appears 235 00:28:20,720 --> 00:28:28,960 that all of the laboratories are experiencing this phenomenon awesome so this is total thc by 236 00:28:28,960 --> 00:28:37,120 laboratory in washington state in 2022 and as you can see this appears to be the flower mode and 237 00:28:37,120 --> 00:28:46,560 this would be the concentrate mode i'll just point out what i find interesting so okay lab 10 lab 11 238 00:28:46,560 --> 00:28:55,760 look quite similar you know lab 2907 is interesting in that they're the only laboratory that has a 239 00:28:55,760 --> 00:29:07,440 a triple mode so they've got a mode around say 50 so maybe they do a lot of teeth or bubble hash 240 00:29:07,440 --> 00:29:15,440 testing that's a possibility i'm just kind of curious then lab 2908 they have a similar 241 00:29:15,440 --> 00:29:23,760 distribution lab 2909 it looks like they do significantly fewer concentrate tests there's 242 00:29:23,760 --> 00:29:30,800 only a small tail down here nothing's actually jumping out at me about 2911 those are pretty 243 00:29:30,800 --> 00:29:37,520 normal distributions right and they're they're different right and so these would be the different 244 00:29:37,520 --> 00:29:43,040 this is actually a good example of two different normal distributions with different moments if you 245 00:29:43,040 --> 00:29:48,640 cut this into two different distributions you know this distribution may have a larger variance 246 00:29:48,640 --> 00:29:55,120 it's going to have what's called greater kurtosis so fatter tails probably the thing that jumps out 247 00:29:55,120 --> 00:30:04,320 to me about 2912 is that maybe their mean of concentrates may be slightly higher but as we 248 00:30:04,320 --> 00:30:10,960 said there's nothing really consequential about having a different mean there could be perfectly 249 00:30:10,960 --> 00:30:17,440 normal underlying explanations so for example perhaps there's a distillate maker and they 250 00:30:17,440 --> 00:30:24,160 simply love to use this laboratory so that could easily just explain this concentrate tail whereas 251 00:30:24,880 --> 00:30:32,080 the bubble hash makers may prefer this laboratory so there's just different underlying causes there 252 00:30:32,080 --> 00:30:39,360 the final thing i wanted to point out here we're not like pointing fingers but i'm just explaining 253 00:30:39,360 --> 00:30:46,560 what cultivators may look at and this may drive market dynamics as we discussed last time okay 254 00:30:46,560 --> 00:30:56,240 so you've got two laboratories here 2913 2914 well you can tell from these distributions are 2913 255 00:30:56,880 --> 00:31:06,880 appears to have a wider variance in flower and their mean appears to be slightly lower than the 256 00:31:06,880 --> 00:31:15,840 mean of 2914 so if you were a cultivator and you saw this data right you sent in your samples to 257 00:31:15,840 --> 00:31:21,680 all the different laboratories well at the end of the day you're trying to sell your cannabis flower 258 00:31:21,680 --> 00:31:28,000 and if you've known they may have noticed that cannabis with the higher thc sells at greater rates 259 00:31:28,000 --> 00:31:34,160 and so if they were comparing these two laboratories they would say oh you know this laboratory 260 00:31:34,160 --> 00:31:43,840 looks like on average we may have higher results and we have a lower variance on average you know 261 00:31:43,840 --> 00:31:51,120 and with lower variance our thc may test higher at this laboratory once again there could be other 262 00:31:51,120 --> 00:31:57,040 underlying mechanisms here as we discussed the different clients using the different labs if 263 00:31:57,040 --> 00:32:06,720 that were the case then you may see people leave laboratory 2913 go and use lab 2914 over time more 264 00:32:06,720 --> 00:32:14,480 and more people use lab 2914 and the methods kind of standardized towards this so that's sort of the 265 00:32:15,200 --> 00:32:22,800 concern about lab shopping is that you know any lab with the slightly different variance as we'll 266 00:32:22,800 --> 00:32:29,360 show later these may actually be surprisingly in the same ballpark so as you can see the lab that 267 00:32:29,360 --> 00:32:38,720 tests the most samples you know has a mean of 22.26 and a variance of around 21 so you could do a 268 00:32:38,720 --> 00:32:45,840 difference in means test between all these different laboratories and see if any of them have 269 00:32:45,840 --> 00:32:52,640 you know statistically different means but as I said that's not really going to take you too 270 00:32:52,640 --> 00:32:59,840 too far and in fact there could be perfectly good explanations for this so for example you know lab 271 00:32:59,840 --> 00:33:08,400 11 they're only testing almost 200 samples level of magnitude different so lab 10 their mean is 272 00:33:08,400 --> 00:33:15,200 significantly different you know perhaps they just have an entirely different type of clientele so 273 00:33:15,200 --> 00:33:23,920 that's what's going on with THC just wanted to point out the other compounds for you real quick 274 00:33:23,920 --> 00:33:30,320 because it helps to look at a bunch of different distributions the distribution and CBD by laboratory 275 00:33:30,320 --> 00:33:36,480 2907 has a really really wide distribution so here's an interesting one so this one is 276 00:33:36,480 --> 00:33:43,440 you know not quite a normal or logarithmic distribution if you were this laboratory you 277 00:33:43,440 --> 00:33:48,160 know you may want you may be looking at the other laboratories and think oh you know maybe you know 278 00:33:48,160 --> 00:33:54,560 maybe there's something different going on with our method however lab 2913 also has this interesting 279 00:33:54,560 --> 00:34:00,720 distribution once again this could just be something unique about the clientele of these 280 00:34:00,720 --> 00:34:07,600 two different laboratories and then once again lab 2914 is a really wide distribution of CBD and so 281 00:34:07,600 --> 00:34:15,760 my takeaway from this is that okay let's say you were growing CBD rich flour you may not want to 282 00:34:15,760 --> 00:34:22,880 use you know these laboratories that have a really wide distribution of CBD because it doesn't look 283 00:34:22,880 --> 00:34:29,920 like these may necessarily be accurate say right you've got 10 laboratories and you know eight of 284 00:34:29,920 --> 00:34:37,040 them are measuring in this manner appears that those may be the laboratories that have the the 285 00:34:37,040 --> 00:34:42,240 more precise method i'm not certain what's going on with with those odd distributions so you could 286 00:34:42,240 --> 00:34:49,120 ask the laboratory about it or pick one that has a slightly more normal distribution that's how i 287 00:34:49,120 --> 00:34:57,120 would approach it so here are moisture content values by laboratory and as you can see they're 288 00:34:57,120 --> 00:35:03,360 actually all over the place they're at least different they're quite different distributions 289 00:35:03,360 --> 00:35:08,560 from lab to lab and i think this is something interesting to point out in that we may have 290 00:35:08,560 --> 00:35:17,840 been quite critical about the THC and CBD ratios but everything's relative so we were just looking 291 00:35:17,840 --> 00:35:24,960 at those analytes so yes the distributions may have looked slightly different but they're all 292 00:35:24,960 --> 00:35:31,120 generally in the same ballpark similar looking distributions they started to look a little 293 00:35:31,120 --> 00:35:38,400 different for CBD but good sign for measuring cannabinoids however when you look at say moisture 294 00:35:38,400 --> 00:35:44,000 content it looks like the laboratories are fundamentally just testing differently right 295 00:35:44,000 --> 00:35:53,120 so you've got lab 10 lab 11 lab 2907 is measuring moisture more on this end of the distribution 296 00:35:53,120 --> 00:36:01,120 or or on this end of the scale same for 2908 then lab 2909 you've got this really peculiar 297 00:36:02,080 --> 00:36:11,840 distribution at a different scale really than 2908 lab 2910 looks slightly different than the 298 00:36:11,840 --> 00:36:18,560 the other ones it's the most similar to lab 10 it looks like then lab 2911 is is pretty similar 299 00:36:18,560 --> 00:36:27,760 to 2910 2912 has what appears to be you know the the most normal of distributions however it does 300 00:36:27,760 --> 00:36:37,200 have a you know a skewed tail here that's a little interesting and then 2913 you see slightly a 301 00:36:37,200 --> 00:36:47,360 slightly peculiar distribution and only ever slightly but it's similar i would say to 2912 302 00:36:47,360 --> 00:36:57,280 and that except for this oddity and then 2914 is once again it's kind of similar but it also has 303 00:36:57,280 --> 00:37:03,520 oddities around this end of the distribution what this would indicate to me is you know maybe 304 00:37:03,520 --> 00:37:14,320 there's something going on with labs 2913 and 2914 how their method is testing above 10% moisture 305 00:37:14,320 --> 00:37:20,880 and once again there may not be anything nefarious about this it just maybe they may have their 306 00:37:20,880 --> 00:37:30,000 instruments calibrated at a certain part of the scale and so they may just need to maybe we just 307 00:37:30,000 --> 00:37:38,000 need to look at how our method is testing moisture content above 10% similarly other labs may say oh 308 00:37:38,000 --> 00:37:46,720 maybe we're you know structurally testing moisture differently so maybe the laboratories can come to 309 00:37:46,720 --> 00:37:55,600 a consensus on what the appropriate method to to test moisture is and they can all have their own 310 00:37:55,600 --> 00:38:04,400 proprietary takes but would it hurt for them to at least agree upon you know around like what 311 00:38:04,400 --> 00:38:10,960 scale of this moisture thing should be on because it looks like these labs think moisture should be 312 00:38:10,960 --> 00:38:16,640 measured in this manner and then these labs think that moisture should be measured in this manner 313 00:38:16,640 --> 00:38:23,760 let's just look at the last one just to be complete and that would be water activity and it's measured 314 00:38:23,760 --> 00:38:34,400 in peculiar units on a scale of zero to one it's measured in aw so here is water activity they're 315 00:38:34,400 --> 00:38:40,800 not perfectly normal distributions i was going to say they're not that bizarre um i guess with 316 00:38:40,800 --> 00:38:49,920 with chunks out okay so right off the bat 2907 is slightly lower 2909 has a peculiar double load 317 00:38:49,920 --> 00:38:58,560 2910 that's a pretty normal distribution just really really wide tails high variance probably 318 00:38:58,560 --> 00:39:05,840 high kurtosis nothing wrong with that that may just be how moisture is in cannabis here's 2911 319 00:39:05,840 --> 00:39:16,240 slightly less variance a bit more of a skew 2912 quite skewed and their mean looks more similar to 320 00:39:16,240 --> 00:39:24,240 2907 and then once again 2913 and 2914 are slightly different and so if you look at them as a whole 321 00:39:24,800 --> 00:39:30,720 they're not wildly all over the place but they are slightly different on first sight they appear to 322 00:39:30,720 --> 00:39:38,800 be more different than the thc distributions that's subjective but that's my take once again 323 00:39:38,800 --> 00:39:46,320 you've got to weigh the costs and benefits obviously people care most about the psychoactive compounds 324 00:39:46,320 --> 00:39:52,720 the thc and the cbd but as you pointed out at the beginning you do need to be concerned about the 325 00:39:52,720 --> 00:39:59,680 shelf life of these products right you don't want things molding on the shelf we'll get more into 326 00:40:00,240 --> 00:40:07,040 microbes and micro toxins we have those results so we can look more at those but i think the take 327 00:40:07,040 --> 00:40:14,400 away from today is you know washington state may actually be doing a fairly good job at measuring 328 00:40:14,400 --> 00:40:23,440 thc and cbd and they may simply need to relook at some of the more minor tests like moisture content 329 00:40:23,440 --> 00:40:30,480 and water activity because those are the ones that actually look anomalous versus versus just the the 330 00:40:30,480 --> 00:40:37,920 thc and the cbd you know maybe they can get around to it now that they're stabilizing their 331 00:40:37,920 --> 00:40:43,200 cannabinoid method so that's sort of my big takeaway from the day remember we've seen all the 332 00:40:43,200 --> 00:40:53,280 the peculiar plots and now you see the thc again by laboratory and once again this laboratory has 333 00:40:53,280 --> 00:40:59,040 a peculiar mode in the middle maybe there's a perfectly good explanation for that and then 334 00:40:59,040 --> 00:41:06,080 once again the mode that we're the most concerned about that as far as fraud goes would just be 335 00:41:06,640 --> 00:41:15,440 the thc mode and from looking at all the other modes yes there may be slight variances in in 336 00:41:15,440 --> 00:41:23,200 variance and mean but they're generally in the same ballpark but as we've pointed out small 337 00:41:23,200 --> 00:41:30,160 differences matter i don't think this is the end of this analysis but i think this was a interesting 338 00:41:30,160 --> 00:41:36,400 take on the data well i'll let you in on the future work that needs to be done we're looking 339 00:41:36,400 --> 00:41:43,200 at the traceability data in washington state and so this was our first take on looking at the 340 00:41:43,200 --> 00:41:50,800 laboratory results we still have to get through all the inventory the strains the sales a lot more 341 00:41:50,800 --> 00:41:58,480 data to crunch and as you were pointing out it's awesome having this open data well you need good 342 00:41:58,480 --> 00:42:05,520 people good data scientists to actually crunch the numbers and this is golden data that's just been 343 00:42:05,520 --> 00:42:12,880 sitting around in the open and i haven't seen anyone do a really really thorough analysis of it 344 00:42:12,880 --> 00:42:21,040 and that's exactly why the cannabis data science team and canlittics are here so over the coming 345 00:42:21,040 --> 00:42:28,960 weeks let's put our minds together organize curate and analyze this data because it's just sitting 346 00:42:28,960 --> 00:42:35,840 there right for us i think we can answer many cool interesting questions with it so let's embark i 347 00:42:35,840 --> 00:42:40,240 think it's going to be a lot of fun sounds awesome you know i was just like looking at my whiteboard 348 00:42:40,240 --> 00:42:45,520 because i happen to have like the uh all the fields and files for the erd i was going to do 349 00:42:45,520 --> 00:42:51,840 washington state that's still on my whiteboard it's been there since last winter but you do so 350 00:42:51,840 --> 00:42:56,240 much you know we're good it's so that's great we're coming back to that because there is a lot of 351 00:42:56,240 --> 00:43:03,040 data in washington state there certainly is you know how we are we're the tortoise so it may have 352 00:43:03,040 --> 00:43:10,640 been a year but this data no one else has looked at it so that's sort of what's crazy right it's 353 00:43:10,640 --> 00:43:17,520 been almost a whole year and no one's looked at this data because it's such a daunting task and 354 00:43:17,520 --> 00:43:23,840 in fact i was reading a really interesting article i'll have to share it with you that was making the 355 00:43:23,840 --> 00:43:31,120 point that's where a lot of people stop is they stop in the data curation phase people have really 356 00:43:31,120 --> 00:43:42,320 cool ideas for apps and they're building all this shiny software but a lot of it requires data data 357 00:43:42,320 --> 00:43:48,720 that's been carefully curated and cleaned and that's what us awesome data scientists are here 358 00:43:48,720 --> 00:43:55,680 to do and help out is if we can get this data cleaned and organized then it unlocks an incredible 359 00:43:55,680 --> 00:44:03,680 amount of potential simply a daunting task but it's not hard it's dirty and that's why i like to 360 00:44:03,680 --> 00:44:12,080 compare our work to that of a plumber and that's why you pay a plumber a decent amount it's dirty 361 00:44:12,080 --> 00:44:20,880 unglamorous work that no one wants to do and we'll roll up our sleeve grab the plunger and go at it 362 00:44:20,880 --> 00:44:27,440 so we're going to go grab the data plunger and get everything unclogged and flowing smoothly 363 00:44:27,440 --> 00:44:32,320 and so all this awesome traceability data that's sitting there right in front of us 364 00:44:32,320 --> 00:44:38,640 we'll get that flowing nicely into people's hands it's great and and so thank you thank you thank 365 00:44:38,640 --> 00:44:45,040 you thank you thank you i want to thank you once again for your eyes your ears your brilliant minds 366 00:44:45,040 --> 00:44:50,880 it's you that really moved the ball forward thank you for helping advance cannabis science 367 00:44:50,880 --> 00:45:18,880 could it do it without you thank you