Weekly trends

[jsmaje]

A sufficient number of weeks has now passed since the start of Stardust Top 100 to assess trends in general activity.

A new graph ('Weekly total increase') shows the week-on-week increase in total score of all dusters within a selectable rank range as a percentage of their initial activity between 6-13 Feb 2008 (thus compensating for the disparate absolute scores).

Shown below are the current plots of (a) the present entire range: 1-100, (b) the top 6: 1-6, (c) the remaining 94: 7-100, and (d) those ranked 50-100:



While total activity appears to be roughly steady, it would seem that this is simply due to an assiduous top 6 offsetting a steadily declining contribution from the rest. Last week the lower 50 in fact scored only 27% of their share at the start, and I'd guess there's an even greater decline below the top 100.

There are understandable reasons for this of course: personal burn-out, boredom through lack of novelty (and particularly new movies), and may I venture, loss of interest amongst the scientific team themselves in providing forum feedback (pace Bryan - I know you're doing your best! The occasional blog entries are nevertheless very welcome; more please!).

It will be interesting to see how much things change whenever we get to phase 3.

John

[Star Gezzer]

What category do you fall into John? You have only put up 2 points in the last two weeks. I realize that it is your job to keep everyone straight but come on lets have a little participation here. Your buddy Kevin and I can not do it all. he he Or are you Kevin and busy running your program to stay ahead of me Lets see your stats.

[jsmaje]

What category do you fall into John? You have only put up 2 points in the last two weeks. I realize that it is your job to keep everyone straight but come on lets have a little participation here. Your buddy Kevin and I can not do it all. he he Or are you Kevin and busy running your program to stay ahead of me Lets see your stats.

Scored an own goal there didn't I, being a (d)

My non-contentious point is that those of us below the top 50, let alone top 6, seem to be finding other things to do with our time until perhaps the project (at least our part in it) moves on.

In my case, I admit taking recent time out writing more useless bits of program and making random posts to the forum rather than dusting, since most of the movies presently available seem to be have been done to death, and there are in fact other things to life than Stardust - yeah, I know that's difficult to believe!

And my personal stats aren't anything to be specially proud of:
Cals correct: 6948
Cals incorrect: 45
Total reals viewed: 29756
Rank: 61
Specificity: 99.88%
Sensitivity: 98.83%

Meanwhile, regarding my "buddy Kevin": OK, OK, enough! I admit having also written the program called 'KevinACourtney' ... or maybe not ...

John

[jsmaje]

Last week's Top 100 scores by the middle-rankers (45-70, admittedly including myself) increased by a mere 11% of their contribution in early February, there having been a steady decline over this time:



How much I wonder is due to the alternative lure of theGalaxy Zoo?
And are or should the team be concerned about this at all?

John

[Nikita]

There are many, many personal reasons for non-participation in the project. I doubt that there could be a single cause. The Zoo has been out for quite some time. Perhps some switched over, which would be fine. The amount of effort required for each project is very different. GZ is easier and has much more variety of images to look at. SD required the user to interact with the program and is more challenging. There is yet another one that requires far more than what we do, called systemic. And still there are others. I think that each appeal to different groups of people.

Also, it must be remembered that it is the middle of summer for the ones in the Northern Hemisphere, and they may be involved with outdoor activities. (Spring is coming soon for our Southern Friends!). Last week, America had the 4th of July holiday, and that can be factored in too.

In order to get good data, we would need long term data. Even then, we are looking at a global group, which is inviting an endless variety of variables. Local, regional, national, and international factors then need to be considered. If the overall trend is lack of participation, I think it would be safe to say that people have lost interest, something that is normal and expected for anything long term. It is why all products, events and projects are "New and Improved" or in other ways, advertised differently to the public.

I can't answer if there is a concern at this point or not. Perhaps as 3 comes out, there will be renewed interest. Also, as progress is being made on the samples, the reports will cause more interest. We have so much more to do, perhaps this should be considered a calm before a new wave comes?

[bmendez]

Hi All,

Thanks a lot John for your measurements of the activity level here at Stardust@home. It is very interesting. We recently analyzed some of the participant data from Phase 1 and found that 52% of all the searches performed during that period were done by only 1% of all the dusters. So there is a group of super-dusters who have been doing the majority of the searches.

This isn't all that unexpected. A month ago, at an astronomy conference in St. Louis, I shared these results with a colleague of mine who is on the Galaxy Zoo team. He said that they find pretty much the same thing, a small group of super-volunteers does the majority of the work on the site.

That shouldn't diminsh the contributions that everyone makes to the project. Everyone's contribution is useful. But I am awed by the dedication of those special few.

Now, indeed the level of participation in Stardust@home has dropped off in recent months. We think that is natural. Afterall, there have been no new movies added since the start of Phase 2, and most of those movies have been seen plenty of times.

The Stardust@home team is still working hard, but in less visible ways. Andrew, Zack, and others have been spending lots, and lots of time learning how to extract candidates from the interstellar tray. This is extremely important and delicate work and takes time. It may not be all that flashy, however. Will and Robert have been upgrading hardware and software for the project, and Anna has been preparing movies for the eventual release of Phase 3. The Red Team has been carefully sorting the alpha list of candidates to help prioritize a list for future extractions. And actual analysis of the extracted samples has been underway.

Myself, I've been away on numerous business and personal trips: my grandmother passed away from cancer a short while ago and I was spending as much time with her and the family as I could. I've also been attending lots of conferences and workshops over the past month, developing outreach programs to students, teachers and the public. In fact, I've been trying to develop another citizen science project with the science team of NASA's WISE mission to involve the public in asteroid searches and research. The work you all have done has been quite inspirational to the scientific community. I get questions and praise about Stardust@home at every conference/meeting I attend.

By the end of the summer, we will finally be ready to begin Phase 3. I suspect we will also have some more definitive results on the analysis of the samples extracted so far. The start of Phase 3 and news of our findings will very likely rev up the old Stardust@home engine again. I only hope Robert's new servers are up to the task

Thanks all,
-Bryan

[jsmaje]

Thanks for that feedback Bryan. Please accept my condolences regarding your grandmother, and respect for all the hard work you and the rest of the team are putting in.

There has in fact been a small but noticeable increase in Stardust Top 100enthusiasm since the latest extractions and analyses, despite having proved negative. Whilst I can't say I'm surprised, at least I shan't have to eat my non-existent hat just yet!

Users of the program may have noticed a week's gap in the data until today. The reason is that my computer was attacked by one of the latest and most malicious viruses called "Win AntiVirus 2008". It causes numerous malfunctions, then raises repeated virus alerts and pretends to be (and looks like) a legitimate Windows virus-removal program. Whether you buy it (or if you've got any sense, don't buy), the problems will continue and multiply. It also disables and corrupts the Task Manager, Registry Edit, System Restore, and all other ordinary ways of dealing with such problems. As far as I'm aware, the legitmate ant-virus programs have yet to catch up with it and, after a week of futile effort by my computer experts, it needed a complete re-installation of Windows with all the hassle of then having to re-install all my applications and web-access. You have been forewarned.

Hence, I have had to interpolate an average of individual scores between 23.7.08 and 6.8.08 for the 30.7.08-dated week, so the scores and ranks for the latter are only estimates.

John
PS: I can guarantee that the Stardust Top 100 program itself has always been and remains virus-free!

[jsmaje]

The latest weekly top 100 activity graph:



Predictably I suppose, there was a significant increase around the time that summer holidays finished in the N. Hemisphere (which has the majority of contributors, as was one of Nikita's points), enhanced no doubt by a resurgence of enthusiasm following the initial putative track/particle extractions (as Bryan hoped). Meanwhile, the latter have so far all proved negative, perhaps accounting for the subsequent steady decline from that early September high to the present almost lowest point.

Those who care to experiment with the program (i.e. changing the rank range included) will find that the greatest fluctuations in activity are amongst the top scorers, who will of course have a disproportionate effect on the overall figures.

In this regard, I've always found it surprising that the basic distribution of personal scoring rates - first graph in the Stardust Top 100 program - almost resembles an exponential (or perhaps Poisson(?), as opposed to e.g. linear) distribution. Bryan implies that this has also been the experience of Galaxy Zoo. Is that a common psychological phenomenon, perhaps resulting from the societal distribution of Obsessive-Compulsive Personality Disorder ? Or may it simply be statistically predictable from some other mathematical or competitive-games theory, etc?

There were several suggestions earlier in the project about how useful this project might be as a psychological case study totally independent of its science objective; have any such studies yet been instituted?

John

[pier]

Hi

in my opinion the pattern of the Top 100 scores should be similar to that of the ranking of the New York Marathon, having on the vertical the time spent in the run and on the horizontal the single person.

This kind of graph is not a statistical distribution but just a ranking.

Ciao

Pier

[jsmaje]

In my opinion the pattern of the Top 100 scores should be similar to that of the ranking of the New York Marathon, having on the vertical the time spent in the run and on the horizontal the single person.
This kind of graph is not a statistical distribution but just a ranking.
Pier

Right, but why this consistent and conspicuously concave graph shape (as opposed to linear or conceivably convex)?



Obviously some people will be more enthusiastic or fit enough to engage in any endeavour such as SD@H or the NY Marathon, but I'd have naively thought that would lead to a more-or-less linear spread. I've tried, but can't find any actual NY Marathon graph of this sort - what shape is it? - you give no link. Even then, for comparison purposes, all professionally-trained and would-be athletes would have to be excluded because of the skew they would most likely produce; I'm not aware of there being any equivalent 'professionals' in the same sense amongst ordinary Stardust hunters.

What I'm asking is whether the particular mathematical shape of the Stardust scoring curve is nevertheless predictable, like a Poisson random event curve, from basic 'participation' statistics (if there's such a thing - if not, there's a PhD in it for someone - unless already subsumed by game theory?)

Also whether it may have anything useful to say in psychological terms, such as where enthusiasm potentially melds into rivalry (at all levels within the rankings, not necessarily just the top few), and why, after including all other possible factors involved, the result should then be a systematic flattening-out of the distribution.

John
(sorry to be worrying about such abstruse trivia - just trying to keep my brain alive until phase 3).

[pier]

Hi John,

I found that the pattern of the Top 100 curve is well fitted by

y = a(x - b)^c

where

a = 2229761
b = -5.0745613
c = -1.3463982. (CurveExpert 1.3 download from the webb)

This curve is not statistic. CurveExpert works with Excell; you have to copy the table of the Top 100 (in excell format) in Curve Expert and ask for the best fitting.

Ciao

Pier

[jsmaje]

y = a(x - b)^c

I should have thanked you earlier for that pier, but am still waiting after 3 weeks for an explanation from someone why the top 100 score curve has that particular shape and formula. I'm no mathematician - is that an exponential, Poisson distribution, parabolic or what?

And if your formula is relevant (rather than just being a coincidental match), what might it tell us about human behaviour, including what the variables a, b & c could actually represent?

Surely someone amongst us (including the team) has, or knows someone with, the appropriate mathematical / statistical / psychological expertise to offer an opinion.

Nothing to do with the dust (if there is any) of course .

[DTF]

John,

I'm no expert on this either, but the formula that Pier found merely defines the curve of the graph that was formed and the values of a,b, and c are completely meaningless other than they roughly define the shape and values along this curve. As Pier points out this is not a statistical graphing. It is also not a geometric curve, which is what the formula is defining. The formula is assuming that the number 100 is 100 units of 1, whereas the actual graphing is for members ranking from 1 to 100. Certainly the member who ranks number 100 is not 100 times the member who ranks number 1 in value (and certainly not in scoring either directly, indirectly, geometrically, or exponentially).

[jsmaje]

DTF, I understand what you say about the limitations of a purely mathematical description of the score vs. rank graph in the form that Pier has provided.

That's why I previously raised the question of whether or not there may be a particular area of psychological/behavioural analysis (perhaps called 'participation' or 'competitive' statistics) which could nevertheless logically account for such a shape.

Though initially wary, I now like Pier's suggestion that the NY marathon results could be a good comparison, given the mix of participants who may be eager+fast / eager+slow or intermittent / interested / not-specially-bothered, & competitive/non-competitive at whatever level, etc. I've tried but can't find such a graph (nor for the London marathon). They must be available somewhere.

I'm perfectly aware that the project itself and its results depend in no way on such matters, but it was the team who consciously (and understandably) instituted the scoring/ranking scheme as a psychological recruitment and retention strategy.

John

[DTF]

John,

Forgive me if I'm wrong about this, but my understanding was that the scoring system was originally implemented in order to place greater mathematical importance to the members identifying tracks in actual movies who were more acurate in identifying tracks in test movies. I, therefore, tend to think that the ranking of members came more from the recognition of its importance to members participating in the forum, rather than a strategy of "psychological recruitment and retention."