Looking great :D

oh, absolutely. And, well, eliminated is as good a word as culled, lol I did consider several possibilities for handling it, but none of them would be able to produce the level of responses sought in order to make the sample effective. Had the final sample only had 400 or so repsonses, it wouldn't have been an effective survey, and the value of it would be much more limited. The "magic number" for a good base really is 600, and this one squeaked by in the end, lol.

That's actually not possible to do, Tunesy, in an open system. Think about it for a second. Unless you live in a police state, there is always a voluntary aspect to the poll, because the individual must choose to participate or not. The actual issue isn't "picks himself" -- it's how he ends up recieving that invitation to do so. Most scientific polls use an elimination algorithm on the final data gathered. What this means is that basically they would take the data offered above and sort it according to predefined criteria gathered primarily from previous polls and actually discard answers based on key demographic aspects (specifically: gender, income, zip code, and certain variable preferences). This algorithm utilizes a random generation rule as well, and, these days, is preprogrammed into the system for certain things such as phone surveys. Note the key there is the source of that algorithm, and the way it is implemented: after the collection process. The typical survey you see as a pretty snapshot in usa today with 500 respondents is actually pulled from a typical sampling of persons known to be willing to answer surveys. Usually, 2300 to 3100 are sampled, and then the algorithm is applied to reduce it down. Now, excepting the algorithm (which is an analysis level tool, not a data collecting tool) the methodology for getting the data is the same: the option is placed out there and the individuals choose to participate or not. In order to be sure the sample pool was wide enough, the survey was advertised across a great many sites and in several fields (the one I missed was education, which I need to find a method for getting in the next one). Language was a critical consideration as well (I had a spanish translation for it as well, but was unable to get it up and presented in time, so had to settle for only the german and French ones, and wasn't able to get a japanese one). The main impteus for the survey were newsletters and front page notices ont he major poser sites (which does skew things towards this overal community, it was intended as a community survey, so the effort was worthwhile). This ensured that the survey's reach went beyond the forums themselves and into the wider community that doesn't pay attention to the mad ramblings of us all. In order for the community as a whole to be able to be sampled randomly as that link describes it (and I know exactly what the point is -- one of my clients is a member of the NCPP that purchased and expanded on my previous poll from a few years back), there would need to essenitally be a single, combined database of all the various members of the community from which random samplings could be taken. That database would require basic demographic information already in it, or the culling algorithm would be needed to be used after the fact in order to reach the sampling rate specified. Establishing that database would require effort that's simply not possible: no site at present that I know can or would be willing to share their user information in order to form it, and I wasn't able to create a specific site for that purpose (registering to be eligible to take the quiz, essentially). There is too much competition between the various sites to allow for that level of cooperation on the scale that was involved here. So let's look at the numbers: I approached in a worldwide parking lot 200,000 people and asked them to answer a survey. Of that 200,000 people, 18,000 or so took a look. of that 18,000, 2100 started to do the survey. of that 2100, 700 or so finished it. That's an awfully random sample. Statistically speaking, it's roughly about as random as being able to randomly sample all of the purchasers across all of the main sites on a random basis -- which is how retail market research is done these days, as it's proven valid time and time again and is growing in popularity even more (and, I should note, from companies predominatly outside the ncpp as they are finding it more cost effective to move it in house). Since any of the sites doing so would be most unlikely to make that information freely available (which was the purpose of doing this and in part why the final results are so large a sample), that isn't an option. My own site -- Odd Ditty Foundry -- doesn't presently have a large enough population sample registered to be effective, and, again, to be honest, had I used it I wouldn't have shared the data. Basically, I did indeed roll up my sleeves. My problem isn't not understanding the nature of how to conduct a survey, but how to manage an online one of this scope with a limited toolset. One benefit of this experience has been that I was able to acquire a much more advanced tool for surveying that is more inline with my original goals and gives me a greater level of flexibility and control over the process and should reduce the difficulties involved in this one and still retain the integrity with which I've tried to operate this one -- and better, long term. It's also shown me a great many of the reasons that such a thing hasn't been done before, and why the community is still suffering from a huge disconnect between perception and reality insofar as the marketplace is concerned. As a community we need to be a little less cutthroat about some things, and a little more about others. Among those things that need to grow are the areas outside the "mainline" area where the "money" is -- particularly if we're going to grow and improve the overall reputation that users of poser have in the wider world. THis survey is one of the ways to do that. I resist going into this level of detail about my methods and motivations usually because, well, It's boring, it's never easily summated, and it means I make long ass posts that rendo always seems to lock on. But I do consider all of those things. I planned this survey for over a year. The questions alone took two months. I began the follow up in December, and made several changes in the time since the survey started, and now even more that I have this improved software for it. The arguments about the validity of the data are mostly off the mark -- they fail to take into account the fullness of the scope of it, or work off of aspects of analysis. I'm not supplying analysis (and I keep kicking myself for the observations I made in the first post), just the raw data. The rest is up to the individuals.

That's an interesting analysis. I disagree. I think it points instead more to a rather large degree of market penetration. Look, say what ya want. I've now been defending this on three sites. I'm tired, and discouraged that something I put a lot of work into is being dissed as if it were some sort of bad thing. So I'm a teensy bit defensive. If you think it's crap, fine. THink that. AIn't nothin I'm gonna type gonna change your mind. There's the data, I'm gonna do another survey again down the road, I think the the data is good, useful, and worthwhile, and I gave it away for free to the community as a whole. No one's saying you have to use it. The information is valuable enough that a few of the larger sites have requested the information. Its something that I felt needed to be done, and the time was right, and I didn't feel that any other site was going to do it and make it publically available. I did something for the community, and I did it the best I could, and I'm going to keep doing it, and getting better, and ya know, at least I did something. Time will tell.

sigh The less than 1% does not prove your point. The number of respondents is greater than the overall number of active Poser/D|S specific posters across the top ten sites. If the only place the survey had been advertised was inthe forums, this may be the case. It was not the only place -- those costs were associated with advertising it. All survey's are self selecting. Yes, all. Survey's are taken strictly on the basis of willingness to do so. Your concern is the random sample being representative of the overall user base. I believe it is, you do not. We both have valid reasons for our beliefs. Only time will say which of us is correct. As a matter of policy, the NCPP also opposes several other forms of polling, and for a very simple reason: the organization exists in order to further the goals of the members, who generally resent online polling. A great many established places do. To an extent, I feel that way as well. However, there's no other means available to do so that I'm aware of that wouldn't involve sizeable investment of the sort that would preclude the release of the data freely and publically. If you have a methodology that you feel is going to be more effective, please feel free to conduct your own survey and release the results publically and freely. If you want the data, have at it. Weight and interpret it as you feel correct. I haven't provided any weighting, haven't given anything other than a couple of basic stats that I pulled from it rather quickly. That's the analysis part of the equation, and is up to the end user of the data. Tunesy -- yes, it was. Won't happen again. I'll see ya all in June, when I do the next one.

Some quick notes: This survey cost 114 dollars. I paid for that. It was advertised through google and msn using fairly common keywords, and the link was passed through legal and advertising circles as well. There were over 18,300 unique visits. There were over 700 responses. Duplicates were culled. There were 2100 attempts to complete the survey. Rosity and animotions featured the survey in their newsletters which have a higher percentage of non-forum readers and greater response -- 60% of the survey responses came after that exposure. There were no active forum posts during those times. The one area I was unable to get the survey to that I know poser is used in is educational circles -- and I wasn't trying for the pre-viz market in general. All effective polls require a base. Regardless of the specific type of survey, all polls currently conducted today are built up on top of and utilize existing data from previously collected sample data -- in some cases stretching back 18 years for marketing, and 3 to 6 for political depending on the nature of the poll. That initial base sample did not exist for this particuar group until now. That initial base sample is always pretty much conducted exactly like this one. That baseline is critical to the developmetn of a core dataset that is then grown and refined. Average sample refinement runs over 6 to 12 runs in order to reach an accuracy level of less than 3%. 3 to 5 runs to get below 5%. The people active in the forums across the ten most frequented sites comprise less than 1% of an estimated 300,000 installed and active community users. You can nit pick and quibble all ya want. I'm known for doing something similar, so I understand the impulse. However, you weren't aware of those points above. I'm going to conduct the next survey in either April or June (I'm still parital to June, but the new software I have for the surveys is really incredible and might speed the process up a great deal). THe survey will be easier to take, but somewhat longer. It won't error out, and there will be additional capabilities for portection of information (it uses a non-identificable cookie so you can return to and complete). Each time I do the survey, I'm going to continue my quest to expand on the dataset and increase the accuracy. That's the way the methodology works, and not only is it sound, but its know to produce more accurate measures. As I've said before -- you'll get the value you want out of it. There's good value there. A lot of it. If you are predisposed to think it valueless, well, ain't nothin I'm gonna say is gonna change your mind, and I've forgotten that a couple times. Its good data. Its public data. I've got four sponsors for the next edition's freebies. I'm looking to get 10 total. They would provide some sort of free item as the incentive and it would be exclusive to the poll. IF anyone is interested, you can contact me through IM or email here. Again -- thanks to everyone who took the survey. Enjoy the data. I'll be back again with more...

Well, won't know for sure until we see P7 come out, but based on the past history of the company (EF's been around a good long time, lol) they keep to that schedule, pausing once every four versions for a major chunk of work.

I'm inclined to think that it's about right for them, and thinking back to some of the goals that the CL team had, it fits well. Oh, and I agree -- it is too short a cycle. But profit cycles generally fade hard if there isn't a new version in roughly that same period, which makes it hard to fund development. Personally, I think a three year cycle is just about right.

Message edited on: 02/14/2006 15:53

incidentally -- arcady -- you got guts, goin' polisci and all :)

If EF's ownership of Poser works out the same as their other programs, then we're going to be looking at an 18 to 24 month cycle for poser. If that's the case, then there will be a new version of poser out 4th quarter this year at the earliest, more likely 1st or second quarter next.

Arcady's means value is incorrect in sampling in this margin. arcady is correct that self selection is not the most desirable method -- however, there is no methodology present that allows any other method. Which is to say that the argument is valid. The view is too narrow, though. The data's valid.

Just to clarify, yet again: THere is no IP information available in the files now. IP information is nt an effective means of establishing identity (if it was, there'd be a lot fewer spammers).

multicore isn't multiprocessor. There are some issues with the way that The intel instruction set (and the code in the OS used to run it) runs on AMD systems. AMD has filed an anti trust lawsuit and the two companies have entered into negotiations in order to solve the hardware aspect of the issue. In most cases, this won't be overtly apparent, but in processor intensive situations, such as 3D and games, it shows up with some interesting odd quirks here and there. Programs which take advantage of Intel's hyperthreading actually function worse and may introduce errors in complex calculations when run under AMD chips -- and the reverse is true as well -- becuase of this issue. In manfuacturing, where an error in a calculation can be the difference between life and death, this becomes a pretty serious issue, and it has affected high end applications already. Ergo -- wait a bit.

forgot the all powerful one, lol!!!! Mutiple Undo!!!

Thanks to svdl, the data is once again available. It is now, however, in access 2003 format, and will require that program in order to read the results. The concerns expressed earier have been addressed.

Well, it might look that way, but that's incorrect. There's a critical response point for polling. That response point is 600. Statistically speaking, a response point of 600 gives you a significance of 1000, meaning that it can be applied, with reasonable certitude to a base of 6 million people. 600 responses is the minimum required by most major polling houses, and in general they cull duplicates from there. Its a funky area, statistical polling -- but yes, John, I can say that the results apply within a roughly 8% variable to the entire community. Can I say what an individual will think? No. THere's no means of doing that. Can I say what the community as a whole will think? Yeah. Human group dynamics does follow known patterns that can be quantified. Your argument seems valid, but it isn't. And I'm not able to effectively refute it without getting into a class on sociological statistics. You can, however, google it out to your hearts content :)