Poll: Did you ever pirate Ableton?
-
- Posts: 102
- Joined: Sat Jun 10, 2006 7:12 pm
- Contact:
What are you talking about? This is just plain wrong. In fact, the data has gotten better since my last review of the statistics. We've come down to a confidence interval of +/- 6.4% with a confidence level of 95%.kovacs wrote:Statistically this poll means zilch because you can't qualify the data - and you have no clue how many people did NOT respond to it.b0unce wrote:I dont get the people who say the poll means nothing...
of the people who took part in the poll 105 people went from crack to license, thats 105 x 450 = 47,250 euro. and thats only the initial price of admission, after that theres upgrades and addons.
Ask someone at ableton HQ if 47,250euro means nothing to them.
This means that out of the registered user population (17,104 now) we can say with 95% confidence that between 41.6% and 54.4% used the crack before they bought Live.
Translating that:
We can say with 95% confidence that, only among forum users, between 7,115 and 9,304 users pirated live before they bought it. If each one eventually paid Ableton $250.00 then between $1.78 million and $2.33 million of Ableton's revenue comes from people on this forum who pirated Ableton first. Cold, hard stats.
If you want to argue these figures, please know the slightest bit about statistics, confidence intervals, confidence levels and the relation to sample size. Nonsense comments like "you can't qualify the data" I would suggest you waste the time of your former math professor with instead of posting here.
The poll is not as accurate as you think.Allison Redhead wrote:What are you talking about? This is just plain wrong. In fact, the data has gotten better since my last review of the statistics. We've come down to a confidence interval of +/- 6.4% with a confidence level of 95%.kovacs wrote:Statistically this poll means zilch because you can't qualify the data - and you have no clue how many people did NOT respond to it.b0unce wrote:I dont get the people who say the poll means nothing...
of the people who took part in the poll 105 people went from crack to license, thats 105 x 450 = 47,250 euro. and thats only the initial price of admission, after that theres upgrades and addons.
Ask someone at ableton HQ if 47,250euro means nothing to them.
This means that out of the registered user population (17,104 now) we can say with 95% confidence that between 41.6% and 54.4% used the crack before they bought Live.
Translating that:
We can say with 95% confidence that, only among forum users, between 7,115 and 9,304 users pirated live before they bought it. If each one eventually paid Ableton $250.00 then between $1.78 million and $2.33 million of Ableton's revenue comes from people on this forum who pirated Ableton first. Cold, hard stats.
If you want to argue these figures, please know the slightest bit about statistics, confidence intervals, confidence levels and the relation to sample size. Nonsense comments like "you can't qualify the data" I would suggest you waste the time of your former math professor with instead of posting here.
There's a disincentive to answering truthfully. You are asking registered forum members to admit to stealing Ableton's software on Ableton's website. While it's probably true that Ableton isn't paying attention to who's voting even the perception of a threat is enough to skew the results. Being the learned statistician that you are you know that even smaller disincentives than this are grounds for invalidation of polling results.
Furthermore, common sense can tell you that the poll is flawed. According to your poll 98% of crack users end up buying the software. That's preposterous. The continuing, non-paying crack users aren't responding. Why on earth would they? You might suggest that crack users don't use the forum but this doesn't make sense either. You don't need a registration number to post and crack users would need the forum info as much if not more than any other user.
The poll is still interesting for what it does accurately provide, namely that the crack convert theory is true to some extent. Personally, I'm surprised by the number of converts. So, that aspect is interesting and useful.
Win XP Pro : SP2 : Dual Opteron 246 : 1GB Ram : MOTU Traveler : Live 7: Cubase 4: Reason 4
-
- Posts: 102
- Joined: Sat Jun 10, 2006 7:12 pm
- Contact:
As there is will all polls asking about deviant behavior, sexual behavior, voting in contentious elections. Etc. etc.ruprecht wrote:The poll is not as accurate as you think.
There's a disincentive to answering truthfully.
I have an MBA with a concentration in Stats and Finance.Being the learned statistician that you are
This would be the case if I were making claims about the number of unreformed crack users. I am not, nor would I try with such a small response rate for that catagory.you know that even smaller disincentives than this are grounds for invalidation of polling results.
No, 98% of forum users at a 95% confidence level. Adding the small sample size (2 users?), however, we boost the confidence interval to almost 20%. I'd never make claims with that large a confidence interval. There's just not enough data. This is where I think we are seeing the self-selection you suggest insofar as unreformed crackers just aren't responding to the poll. We could do a cross referencing poll, which perhaps I'll do later. Still, as others have pointed out, this population is probably skewed towards paying users. But that's fine. What's interesting is not the ratio of unrepentant crack users, but rather the number of reformed crack users. That's probably not badly impacted by retribution bias, as many of them have admitted as much in the comments.Furthermore, common sense can tell you that the poll is flawed. According to your poll 98% of crack users end up buying the software.
We agree here.That's preposterous. The continuing, non-paying crack users aren't responding. Why on earth would they?
Yes. I haven't taken anything else from this poll other than there are a buttload of reformed crack users floating around out there. This, I think, is impressive.The poll is still interesting for what it does accurately provide, namely that the crack convert theory is true to some extent. Personally, I'm surprised by the number of converts. So, that aspect is interesting and useful.
What amazes me is the absolute solidity of the ratio. It's been basically 50/50 reformed / totally innocent users for three doublings of sample size. That's just amazingly stable.
Yes, and that bias is found even when the polls are anonymous. This case is worse as it isn't anonymous - it's Ableton's own website and they could track the results. Most importantly the users know this.Allison Redhead wrote:As there is will all polls asking about deviant behavior, sexual behavior, voting in contentious elections. Etc. etc.ruprecht wrote:The poll is not as accurate as you think.
There's a disincentive to answering truthfully.
Actually, you are. When you extrapolate the results and make claims about the forum population at large i.e., "7000-9000 reformed users" what happens to all of the continuing crack users? There are only so many forum members. Extrapolating the results means there are only 145 current crack users in the forum. This can't be right. You can't extrapolate the data in this poll.This would be the case if I were making claims about the number of unreformed crack users. I am not, nor would I try with such a small response rate for that catagory.you know that even smaller disincentives than this are grounds for invalidation of polling results.
Yes, this is astounding. Even the number of converts is interesting. I'm shocked. I'd like to know why they chose to convert. Morals? Stability? Fear?What amazes me is the absolute solidity of the ratio. It's been basically 50/50 reformed / totally innocent users for three doublings of sample size. That's just amazingly stable.
Win XP Pro : SP2 : Dual Opteron 246 : 1GB Ram : MOTU Traveler : Live 7: Cubase 4: Reason 4
-
- Posts: 102
- Joined: Sat Jun 10, 2006 7:12 pm
- Contact:
You were almost there.ruprecht wrote:Actually, you are. When you extrapolate the results and make claims about the forum population at large i.e., "7000-9000 reformed users" what happens to all of the continuing crack users? There are only so many forum members. Extrapolating the results means there are only 136 current crack users in the forum. This can't be right. You can't extrapolate the data in this poll.Allison Redhead wrote:This would be the case if I were making claims about the number of unreformed crack users. I am not, nor would I try with such a small response rate for that catagory.ruprecht wrote:...you know that even smaller disincentives than this are grounds for invalidation of polling results.
You accounted for the confidence interval in the 7000-9000 figure but failed to realize what the confidence interval means when you used the hard number of 136. It means that we are 95% sure about 136. The remaining 3000 or so users fall into a sort of "we have an idea, but we aren't sure" limbo catagory. If I reduced the confidence level down to 75% we'd get most of them. That's not good stats, however. A 95% confidence level is about standard for research. Discounting the entire poll because of that is silly. That's like saying that because Independents represent 1% of the poll response from Gallup the results for Democrats and Republicans are totally invalid. This is, of course, just not so, even though Independents historically don't like responding to polls. (Several studies on this). Same exact problem here, probably.
If you do the math you will find that of 17,000 some registered forum users you have room to find that with a 95% confidence level as low as 7000 are reformed crack users, 7000 are total crack virgins and the remainder (17,000 - 14,000 = 3000) are divided up this way:
Between 2 and 2900 or so are unreformed crack users and the remainder are likely divided 50/50 between the other catagories. This is because your confidence interval gets really large with small samples. In this case it's nearly 20% (this is huge). We just don't have enough data to pin down these users. These users aren't GONE, they just are way outside the 95% confidence interval so we can't say much about them with confidence. If we were doing a course in survey design we would attempt to find variables that would pull these out. Still, it's beyond our scope here.
This is, statistically anyhow, actually a really good survey. I can boost the confidence level to 99% and still get a 11% confidence interval. Too big for real work, but interesting.
That's my next survey.Yes, this is astounding. Even the number of converts is interesting. I'm shocked. I'd like to know why they chose to convert. Morals? Stability? Fear?What amazes me is the absolute solidity of the ratio. It's been basically 50/50 reformed / totally innocent users for three doublings of sample size. That's just amazingly stable.
![Wink ;)](./images/smilies/icon_wink.gif)
The poll is not useless. The results themselves are interesting. But, you cannot extrapolate the data and apply it to the forum at large. Playing with confidence levels and intervals cannot rectify the flaw that prevents extrapolation - namely an unquantifiable disincentive that applies to only one answer choice.
You simply do not know or have any clue how many die-hard pirates are in the forum population. The two pirate responses don't help you because we don't know how much that number is being affected by the disincentive that does not apply to the other questions. You only know about the people who have answered. The next person to answer could be the last of the legit users and the rest of the forum members are all pirates. You have no way of knowing or even predicting this. That's why you cannot say there are 7000-9000 crack converts in the forum with 95% certainty.
However, I have a solution for you. If you post a similar poll in an independent forum where the disincentive can be brought down to a reasonable level similar to that of any deviant behavior question you'll get more accurate results. A good choice would be http://www.kvraudio.com. You could ask about all hosts in general. You'll get a larger sampling and people will be less concerned with KVR tracking them.
You simply do not know or have any clue how many die-hard pirates are in the forum population. The two pirate responses don't help you because we don't know how much that number is being affected by the disincentive that does not apply to the other questions. You only know about the people who have answered. The next person to answer could be the last of the legit users and the rest of the forum members are all pirates. You have no way of knowing or even predicting this. That's why you cannot say there are 7000-9000 crack converts in the forum with 95% certainty.
However, I have a solution for you. If you post a similar poll in an independent forum where the disincentive can be brought down to a reasonable level similar to that of any deviant behavior question you'll get more accurate results. A good choice would be http://www.kvraudio.com. You could ask about all hosts in general. You'll get a larger sampling and people will be less concerned with KVR tracking them.
Win XP Pro : SP2 : Dual Opteron 246 : 1GB Ram : MOTU Traveler : Live 7: Cubase 4: Reason 4
-
- Posts: 102
- Joined: Sat Jun 10, 2006 7:12 pm
- Contact:
Which we have removed from the analysis.ruprecht wrote:The poll is not useless. The results themselves are interesting. But, you cannot extrapolate the data and apply it to the forum at large. Playing with confidence levels and intervals cannot rectify the flaw that prevents extrapolation - namely an unquantifiable disincentive that applies to only one answer choice.
I admit as much. So? Doesn't really impact the other data.You simply do not know or have any clue how many die-hard pirates are in the forum population.
*sigh* This is simply not so.The two pirate responses don't help you because we don't know how much that number is being affected by the disincentive that does not apply to the other questions. You only know about the people who have answered. The next person to answer could be the last of the legit users and the rest of the forum members are all pirates. You have no way of knowing or even predicting this. That's why you cannot say there are 7000-9000 crack converts in the forum with 95% certainty.
-
- Posts: 102
- Joined: Sat Jun 10, 2006 7:12 pm
- Contact:
That's what gets people in trouble. Using personal judgement to contest the numbers and labeling it "common sense." It's neither common nor sense. We start doing this and then we start looking like American politicians. No thanks.ruprecht wrote:*sigh* put the statistics textbook down for a second and use common sense.
Since the catagories are exclusive (this was intentional) you can happily remove an independent variable or two and still get solid results. Plain stats. Note that I put in a silly option too as a control for "tricksters." That wasn't an accident. It gives us an interesting gauge too. Not many non-serious poll responders. If I had an hour I'd run the analysis on this but we'd almost certainly get a low 1-2% "trickster" figure. In short, this is a pretty damn good sample.
Look, I'll help you:
The best way to argue against the survey is that the sample is non-random. Since we've confined our analysis to the population of the forum you then have to say the sample is a non-random selection of forum users. This means that to make the case that FEWER than 7000 forum users are actually pirate converts you have to somehow show a connection between reformed pirates and a propensity to answer forum polls. Perhaps reformed pirates are more likely to read the General forum? Perhaps they are more likely to read the forum in the fall/winter? Who knows. We don't have any data like this. Even if we did we'd have to show a rather large bias. Let's say there is a bias error coefficient of .75 (this would be outlandishly huge for a bias we can't even quantify right now, and huge for a bias period). That still leaves us with 1,750 users who are reformed pirates or $435,000 or so in revenue @ $250 a pop (which is probably low). This is with beyond worst case assumptions too.
The bottom line is that there are pretty clearly millions of dollars of revenue out there for Ableton that come from reformed pirates. You have to torture the facts pretty seriously (I mean REALLY seriously) to come out with a number under half a million dollars JUST FOR THIS FORUM. I think that's worth noting.
Anyhow, people who have some personal/political reason to deny these figures will always find a way to do so. Inconveniences as small as facts and math won't get in their way.
- Allison
I'm beginning to think you like to argue. I've stated several times I think the poll is interesting and says alot about piracy and the tendency for people to convert to legitimate use. Please don't insinuate that I have some political agenda because you cannot otherwise counter criticism.
I'm simply pointing out that extrapolating the poll data is leading you to incorrect conclusions for the reasons I've clearly explained. I'm sorry you are unable to understand this. It is correct. Maybe one of your professors could explain it further.
I think you're actually on to something very interesting here. I hope your myopia doesn't prevent you from developing it properly.
I'm simply pointing out that extrapolating the poll data is leading you to incorrect conclusions for the reasons I've clearly explained. I'm sorry you are unable to understand this. It is correct. Maybe one of your professors could explain it further.
I think you're actually on to something very interesting here. I hope your myopia doesn't prevent you from developing it properly.
Win XP Pro : SP2 : Dual Opteron 246 : 1GB Ram : MOTU Traveler : Live 7: Cubase 4: Reason 4
-
- Posts: 102
- Joined: Sat Jun 10, 2006 7:12 pm
- Contact:
What was yer first clue?ruprecht wrote:I'm beginning to think you like to argue.
![Smile :)](./images/smilies/icon_smile.gif)
I'm sorry to differ with you.I'm simply pointing out that extrapolating the poll data is leading you to incorrect conclusions for the reasons I've clearly explained. I'm sorry you are unable to understand this. It is correct. Maybe one of your professors could explain it further.
I understand it fine and I just disagree with your reasons for discounting the poll. I've provided data and math to support my case.
Awwww, shucks.I think you're actually on to something very interesting here. I hope your myopia doesn't prevent you from developing it properly.
- Allison