5 ms·

Working on this project with a friend. Any suggestions?

by coryl 16y ago

jRoden 16y ago
The problem with compiling an overall $/oz. amount is that, generally, prices go down as quantity goes up. The data will be skewed if a lot of people are reporting $50 eighths (400/oz.) and the price for an ounce from the same dealer is, say, $300. Maybe track the different amounts separately?
- coryl 16y ago
  Exactly, we aren't incorporating the volume discount when we scale up smaller quantities of weed. This is because we really just don't know what the average margins of sales between different quantities. We could display the averages in each quantity, but it would lack enough sample size. Plus having one nice round number is easier to visualize. I think the final .oz average probably within +/-10% accuracy.
  - dgc 16y ago
    publishing the standard deviation on the normalized price would go a long way towards making the numbers more useful. If you get enough samples you can do a simple fit to attempt to quantify the discount rate.
lmkg 16y ago
Data scrubbing. For example: California: High Quality: $359.19 / oz. (n=72) Medium Quality: $300 / oz. (n=17) Low Quality: $648 / oz. (n=2) <-- ??? The low quality stuff is worth more than twice as much as the medium? The crazy numbers for low quality are on account of one wonky data point ($976/oz). There were also some obviously bunk ones like $1/oz, which someone probably did pay from a friend, but that doesn't reflect street value. It's hard to do this in an automated fashion, though, at least until you have a good amount of data. A good place to start would be automatically flagging figures more than 2 standard deviations from the (local) mean, and hand-curating those. Keep track of "bad" data points, and as you find patterns you can add filters.
- coryl 16y ago
  Yep, we haven't gotten around to filtering and cleaning up entries, but we'll definitely look to eliminate big outliers from our average calculations. A lot of people have been pointing on the same thing you've been pointing out lol. That's why we added the n= for the time being.
- Alex3917 16y ago
  The price for the High Quality doesn't make much sense either even with tons of data points. According to Yelp the most expensive dispensary in SF is Divinity Tree, and I think their onces are $360. Considering that basically everywhere else in the state should be cheaper than this I don't see how the average could possibly be 360.
  - coryl 16y ago
    Not everyone can get licenses to purchase from dispensaries, so a dispensary's price is not necessarily market price. Theres still a "black" market so to speak; of people who buy it illegally.
    - Alex3917 16y ago
      Yes, but the black market price is usually cheaper than the dispensary price.
      - coryl 16y ago
        Not entirely true. If I can't get a license, I have to buy off the black market; meaning I have to go through dealers who traffic under the risk of arrest and prosecution. That creates the premium price that drugs have today, everywhere. Thus it makes sense for a drug dealer to charge more than the price available legally, because buyers without licenses have no choice. However, there are also demand and supply side things going on that we might not know about that is affecting the price.
        jrockway 16y ago
evgen 16y ago
I think one problem you will run into is that you are attempting to compare prices while also keeping fuzzy distinctions like quality. Alabama has cheaper "high quality" weed than California? That tells me that what would sell for $400/oz in CA is not available at any price in AL. If you cut out the user-supplied quality metric or figure out some way to try to normalize this across the samples (e.g. ask submitters if they have purchased in other states and how they would rank this purchase against that past purchase) you would probably get better useful data.
- coryl 16y ago
  Indeed, quality is highly subjective and relative to product available on the market. I have a personal hypothesis that most average smokers can't distinguish high quality vs medium quality. I'm not sure how practical it is to get users to try and compare product quality from different locations; the sample size would probably be very low because most people buy and consume where they live.
  - evgen 16y ago
    True. I was mostly suggesting that you should drop the quality distinction but trying to leave you an out in case there was some clever way to normalize this. I think that in general people would find an accurate average more useful because then they could perform any quality discounting on their own using the known basepoint.
  - dgc 16y ago
    I actually think there could be some objective quality measures, but you would need someone with a fair range of experience to judge. Perhaps number of buds per ounce in the mix or a simple color/image chart to select from?
  - c23gooey 16y ago
    a simpler distinction would be whether the marijuana is outdoor or hydroponically grown
sv123 16y ago
I only ever buy a single gram, but can't put that in. Also what if vendors start putting prices they pay growers, and dispensaries enter prices from vendors, and customers use prices from dispensaries... The same batch of herb will be many different prices, possibly need some way to distinguish a "type" of price that is being entered. That would make the data more valuable also for dispensaries shopping around different vendors, etc. Of course that is sticking to the "legal" weed market.

Working on this project with a friend. Any suggestions?

by coryl 16y ago

Working on this project with a friend. Any suggestions?

The problem with compiling an overall $/oz. amount is that, generally, prices go down as quantity goes up. The data will be skewed if a lot of people are reporting $50 eighths (400/oz.) and the price for an ounce from the same dealer is, say, $300. Maybe track the different amounts separately?

coryl 16y ago
Exactly, we aren't incorporating the volume discount when we scale up smaller quantities of weed. This is because we really just don't know what the average margins of sales between different quantities. We could display the averages in each quantity, but it would lack enough sample size. Plus having one nice round number is easier to visualize. I think the final .oz average probably within +/-10% accuracy.
- dgc 16y ago
  publishing the standard deviation on the normalized price would go a long way towards making the numbers more useful. If you get enough samples you can do a simple fit to attempt to quantify the discount rate.

lmkg 16y ago

Data scrubbing. For example: California: High Quality: $359.19 / oz. (n=72) Medium Quality: $300 / oz. (n=17) Low Quality: $648 / oz. (n=2) <-- ??? The low quality stuff is worth more than twice as much as the medium? The crazy numbers for low quality are on account of one wonky data point ($976/oz). There were also some obviously bunk ones like $1/oz, which someone probably did pay from a friend, but that doesn't reflect street value. It's hard to do this in an automated fashion, though, at least until you have a good amount of data. A good place to start would be automatically flagging figures more than 2 standard deviations from the (local) mean, and hand-curating those. Keep track of "bad" data points, and as you find patterns you can add filters.

coryl 16y ago
Yep, we haven't gotten around to filtering and cleaning up entries, but we'll definitely look to eliminate big outliers from our average calculations. A lot of people have been pointing on the same thing you've been pointing out lol. That's why we added the n= for the time being.
Alex3917 16y ago
The price for the High Quality doesn't make much sense either even with tons of data points. According to Yelp the most expensive dispensary in SF is Divinity Tree, and I think their onces are $360. Considering that basically everywhere else in the state should be cheaper than this I don't see how the average could possibly be 360.
- coryl 16y ago
  Not everyone can get licenses to purchase from dispensaries, so a dispensary's price is not necessarily market price. Theres still a "black" market so to speak; of people who buy it illegally.
  - Alex3917 16y ago
    Yes, but the black market price is usually cheaper than the dispensary price.
    - coryl 16y ago
      Not entirely true. If I can't get a license, I have to buy off the black market; meaning I have to go through dealers who traffic under the risk of arrest and prosecution. That creates the premium price that drugs have today, everywhere. Thus it makes sense for a drug dealer to charge more than the price available legally, because buyers without licenses have no choice. However, there are also demand and supply side things going on that we might not know about that is affecting the price.
      - jrockway 16y ago

evgen 16y ago

I think one problem you will run into is that you are attempting to compare prices while also keeping fuzzy distinctions like quality. Alabama has cheaper "high quality" weed than California? That tells me that what would sell for $400/oz in CA is not available at any price in AL. If you cut out the user-supplied quality metric or figure out some way to try to normalize this across the samples (e.g. ask submitters if they have purchased in other states and how they would rank this purchase against that past purchase) you would probably get better useful data.

coryl 16y ago
Indeed, quality is highly subjective and relative to product available on the market. I have a personal hypothesis that most average smokers can't distinguish high quality vs medium quality. I'm not sure how practical it is to get users to try and compare product quality from different locations; the sample size would probably be very low because most people buy and consume where they live.
- evgen 16y ago
  True. I was mostly suggesting that you should drop the quality distinction but trying to leave you an out in case there was some clever way to normalize this. I think that in general people would find an accurate average more useful because then they could perform any quality discounting on their own using the known basepoint.
- dgc 16y ago
  I actually think there could be some objective quality measures, but you would need someone with a fair range of experience to judge. Perhaps number of buds per ounce in the mix or a simple color/image chart to select from?
- c23gooey 16y ago
  a simpler distinction would be whether the marijuana is outdoor or hydroponically grown

sv123 16y ago

I only ever buy a single gram, but can't put that in. Also what if vendors start putting prices they pay growers, and dispensaries enter prices from vendors, and customers use prices from dispensaries... The same batch of herb will be many different prices, possibly need some way to distinguish a "type" of price that is being entered. That would make the data more valuable also for dispensaries shopping around different vendors, etc. Of course that is sticking to the "legal" weed market.